Anatomical Evidence for a Uniquely Human Depressor Anguli Oris and a Novel Helplessness Signaling Hypothesis Authors: Liat Rotenstreich, Aya Vituri, Moni Shahar, Kfir Eliyahu, Ori Farkas, Nir Edery, Lior Moss, Assaf Marom Published: 2026 ABSTRACT ABSTRACTThe downward pulling of the mouth corner in humans is attributed to the depressor anguli oris (DAO), yet the evolutionary history of this muscle remains unresolved. Classical comparative accounts have often assumed that nonhuman primates (NHPs) possess a homologous lower component within the caninus–triangularis complex (CTC), but this proposition has not been systematically tested. Here, we examined the architecture of the CTC across 10 primate species using a multimodal anatomical approach combining gross dissection, morphometric analysis, histology, and fiber‐level quantification. We analyzed 30 NHP specimens and 61 human specimens to determine whether NHPs possess a discrete mandibular muscle homologous to the human DAO. In all human specimens, the CTC consisted of two distinct bellies, the levator anguli oris and DAO, whereas in NHPs, including great apes, the CTC appeared as a continuous sheet lacking a discrete mandibular DAO homolog. Morphometric and fiber‐level analyses supported this distinction, revealing a characteristic human pattern around the labial commissure consistent with bipartite organization. Although the present study does not test behavioral function directly, it motivates the hypothesis that this derived muscular architecture may have been recruited into an expressive facial signal associated with helplessness during infancy in the context of cooperative breeding. Introduction Facial expression is a defining feature of human communication, enabling individuals to convey emotional states, elicit responses, and coordinate social behavior with striking immediacy [1, 2, 3, 4, 5]. Additionally, facial expressions are not only displays of affect but also socially relevant signals that can help observers predict forthcoming behavior and adjust their own responses accordingly. From an evolutionary perspective, the ability to extract information about the emotions, motivations, or likely actions of others from facial displays may confer important fitness benefits in social decision‐making and interactional success [6, 7, 8, 9]. For over 150 years, evolutionary theorists have debated the extent to which this capacity is shared with other primates [1, 10, 11, 12, 13, 14]. Darwin [1] famously argued for a continuity in emotional expression across species, grounding his theory in homologous muscular anatomy and instinctive responses. By contrast, the downward pulling of the mouth corners, long regarded as the hallmark of sadness [1, 6, 15, 16, 17] and a constituent of other expressions [2, 3], appears to be a human‐specific configuration rather than a shared primate one. This display is prominently expressed in human fetuses [18], neonates, and young children [19, 20, 21], yet appears to be absent from the expressive repertoire of other primates [22]. This distinctive movement, which Darwin highlighted as integral to the cry face [1], plays a pivotal role in allowing a prelinguistic helpless infant to visually signal need and elicit caregiving [6, 23]. When absent, as in infants with congenital depressor anguli oris (DAO) aplasia [24] or Moebius Syndrome [25], the characteristic cry‐face is markedly diminished or absent. In humans, the vertical position of the labial commissure is governed by two anatomically distinct muscles: the levator anguli oris (LAO), originating from the maxillary canine fossa, and the DAO, which originates from the mandible [26, 27]. These opposing vectors converge at the modiolus, enabling fine modulation of mouth‐corner movements. In nonhuman primates (NHPs), homologous structures have traditionally been referred to as the caninus and triangularis muscles [28, 29, 30, 31, 32, 33, 34]. Figure 1 provides an overview of these anatomical components in humans and other primates. Scheme of perioral musculature in primates and humans. In humans (right panel), the levator anguli oris and depressor anguli oris are fully differentiated muscles that converge at a distinct fibromuscular node—the modiolus—located at the level of the labial commissure (white asterisk). The depressor anguli oris muscle originates from the mandible, enabling effective depression of the mouth corner. In nonhuman primates (left panel), the homologous muscle typically terminates at the level of the labial commissure. In great apes (middle panel), the muscle extends further inferiorly but still falls short of mandibular insertion. The isolated muscles depicted below each schematic highlight the evolutionary question addressed in this study. D5 labels the position of the mouth corner. Classical anatomical descriptions often treat the NHP caninus and triangularis muscles as a continuum, collectively termed the caninus–triangularis complex (CTC) [33, 35, 36], based on the assumption that a discrete triangularis muscle, homologous to the human DAO, is present. Yet despite frequent mention of the DAO in comparative anatomical studies [13, 34, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48], its evolutionary origin has remained contested, owing to longstanding disagreements in classical anatomical literature concerning its presence in NHP faces [28, 29, 30, 31, 33, 37, 38]. As early as the 19th century, Scottish anatomist Charles Bell proposed that the DAO was a uniquely human muscle, one he could not identify in other animals and which he believed served no other function but expression [49]. Yet Bell's hypothesis has never been systematically tested. Although classical anatomical examinations have touched upon aspects of facial muscle variation that relate indirectly to his claim [16, 28, 29, 50, 51], it remains unaddressed in a comparative evolutionary framework. At the same time, behavioral systems such as the Facial Action Coding System (FACS) have sought to standardize facial expressions across species [52], further highlighting the need for anatomical validation of expressive homologies. Within this system, Action Unit 15 (AU15), corresponding to the activation of the DAO, plays a central role in signaling sadness or distress in humans [16, 53]. Notably, AU15 has never been documented in any NHP species using FACS adaptations [54, 55, 56, 57, 58, 59, 60, 61], raising the possibility that both the DAO and the expressive configuration it enables are uniquely human. This further suggests that the downward mouth‐corner movement associated with sadness lacks evolutionary continuity across primate species [62]. Aims and Hypothesis Although the human face features a distinct LAO and DAO, comparative studies have debated this bipartite organization in NHPs. The present study aims to examine whether the CTC in NHPs is indeed composed of two distinct muscular units as in humans, including a homolog to the human DAO. Within the evolutionary methodological framework of this study, we refer to the vertically oriented muscle spanning from the canine fossa to the lower face as the CTC in both humans and NHPs, consistent with terminology in classical literature. In humans, this complex consists of the distinct LAO (upper component) and DAO (lower component). Accordingly, our null hypothesis (H0) posits that the CTC in NHPs is similarly composed of two distinct muscle bellies. Our alternative hypothesis (H1) is that the CTC in NHPs consists of a single muscle with no discrete mandibular component homologous to the human DAO. Methods Ethical Statement Cadaveric head specimens of NHPs were obtained from multiple sources: the Zoological Center Ramat Gan (Israel), the Edmund and Lily Safra Center for Brain Sciences (The Hebrew University of Jerusalem), and the Kimron Veterinary Institute (Ministry of Agriculture, Bet Dagan, Israel). Approval from the Institutional Animal Care and Use Committee (IACUC) was not required, as all specimens were collected post‐mortem following natural causes of death. Formalin‐fixed human head specimens were obtained from the Farkas Family Center for Anatomical Research and Education (Rappaport Faculty of Medicine, Technion—Israel Institute of Technology), in compliance with institutional ethical regulations. Samples Specimens from the following primate families were examined: Callitrichidae (Callithrix jacchus), Cercopithecidae (Colobus satanas, Macaca nigra, Macaca silenus, Mandrillus sphinx, and Papio hamadryas), Hominidae (Homo sapiens, Gorilla gorilla, and Pan troglodytes), and Lemuridae (Lemur catta). A total of 30 adult NHP specimens and 61 adult human specimens were analyzed. The species included in the present study and the corresponding sample sizes are summarized in Table S1. The number of available great ape specimens was limited due to the exceptional difficulty to obtain such samples. Gross Anatomy NHP heads were disarticulated at the level of C1 vertebra and fixed in 4% formaldehyde for at least 4 weeks prior to dissection. Human specimens were similarly fixed. A detailed macroscopic evaluation of the orofacial musculature was performed, focusing on the CTC. Careful anatomical dissection was applied to both sides of the face. Small specimens (e.g., C. jacchus and L. catta) were dissected using a Nikon SMZ25 stereomicroscope to ensure precise identification and separation of muscle fibers and adnexa. A pre‐auricular incision was made and extended inferiorly toward the submandibular region. The skin was reflected to expose the platysma muscle at the lower lateral aspect of the face. The vertically oriented fibers of the triangularis part (lower component) of the CTC were identified posterolateral to the labial commissure. In the upper lip region, the apron‐like musculotendinous portion of the quadratus labii superioris muscle was exposed superficially and reflected superiorly to reveal the canine fossa and the caninus part (upper component) of the CTC at its origin. Figure 2 includes representative dissected craniofacial specimens included in the study. Representative in situ dissections of the caninus–triangularis complex (CTC) across primates and humans. Images show dissected muscles prior to removal, highlighting their anatomical configuration and orientation relative to surrounding structures. In all panels, the labial commissure is marked by a yellow, dashed line; muscle fibers above this level (orange arrowhead) correspond to the caninus/levator anguli oris (LAO), and fibers below this level (black arrowhead) correspond to the triangularis/depressor anguli oris (DAO). Together, these fibers form the CTC. (A–C) G. gorilla: (A) Superficial dissection after skin removal. Fibers lateral to the commissure (black arrowhead) could be mistaken for a human DAO, but (B) reflection shows these fibers lie superficial to the platysma and lack mandibular attachment. (C) After reflecting the aponeurosis of the nasolabial muscles (black asterisk in panel B), caninus fibers (orange arrowhead) and triangularis fibers (black arrowhead) are revealed as continuous with one another, without convergence toward the modiolus. (D–E) P. troglodytes: (D) Superficial dissection after skin removal and reflection of the nasolabial aponeurosis (black asterisk). Caninus (orange arrowhead) and triangularis (black arrowhead) fibers are continuous across the labial commissure. (E) Reflection of the triangularis fibers shows they lie superficial to the platysma but again lack mandibular attachment. (F–M) Representative dissections of P. hamadryas (F and G), M. sphinx (H and I), M. nigra (J and K), and C. satanas (L and M) all exhibit the same configuration. In each species, the left panel shows superficial dissection after skin removal, where fibers adjacent to the labial commissure (black arrowhead) could be mistaken for a DAO. The right panel demonstrates that, after reflection of the nasolabial aponeurosis (black asterisk), these fibers are revealed as a direct continuation of the caninus (orange arrowhead). In all cases, these fibers terminate at the level of the labial commissure, intermingling with adjacent perioral muscles, without any distinct extension into the lower lip as seen in great apes, and without mandibular attachment—indicating that no true triangularis/DAO is present. (N) H. sapiens: In contrast, the LAO (orange arrowhead) and DAO (black arrowhead) are discrete muscles converging at the modiolus, with the DAO showing a clear mandibular bony origin. The CTC was isolated, and its fibers at the level of the labial commissure were marked using a tissue‐marking pen. The muscle was carefully harvested, and all specimens were photographed using a Canon EOS 90D camera (Figure 3). Next, we performed morphometric measurements and histological analyses (Figure 4) to facilitate muscle fiber quantification. Representative harvested caninus–triangularis complex (CTC) muscles. Shown are intact CTC muscles dissected from craniofacial specimens across primate species. Muscles are annotated according to their orientation relative to the labial commissure (black, dashed line), with fibers above the commissure corresponding to the caninus, and fibers below the commissure representing either a continuation of the caninus or the triangularis, depending on the species. The images highlight how intact muscle bellies were carefully isolated and preserved during dissection, ensuring reliable distinction from adjacent perioral muscles. Segmental sampling strategy for muscle architecture analysis. Macroscopic images of dissected CTC from a human specimen (left) and a nonhuman primate (P. hamadryas, right). Dashed lines indicate the sampling levels (D1–D9). Hematoxylin and eosin‐stained cross‐sections are presented for each level, showing the distribution of muscle fibers used for quantitative analysis. Scale bars: 2 mm (gross anatomy) and 500 µm (histology). Because the central inference of this study depends on distinguishing a discrete mandibular DAO from caninus fibers that may extend below the labial commissure, we sought to avoid classification based on gross anatomical appearance alone. A discrete DAO was identified when a mandibularly anchored inferior component showed a consistent trajectory toward the labial commissure and was histologically supported by reduced fiber continuity and increased intervening connective tissue relative to the upper component, that is, a bipartite architecture (see Figure S1). By contrast, a continuous CTC was defined as a superficial muscular sheet arising from the maxillary canine fossa, lacking a distinct mandibular anchorage or separable inferior belly, with no reduction in fiber continuity at the commissural level (see Figure S1). In such cases, fibers extending below the mouth angle were interpreted as part of the same continuous sheet rather than as a distinct mandibular muscle. Morphometric Measurements Each CTC was sectioned longitudinally at seven anatomically defined levels (Figure 4): Level D1 corresponded to the superior margin of the muscle at its origin in the canine fossa and D9 to its inferior margin. The labial commissure was designated as D5. Levels D3 and D7 represented the midpoints of the upper and lower portions, respectively. In addition, tissue was sampled at 15% above and below D5, capturing the region surrounding the labial commissure. Muscle width was measured at each of the seven levels using the Measure Tool in ImageJ software. Muscle length was assessed using two measurements: L1, representing the distance from the upper margin (D1) to the labial commissure (D5), and L2, representing the distance from the labial commissure (D5) to the lower margin of the muscle (D9). To distinguish the CTC architecture in NHPs from that of the human face, where it comprises upper (LAO) and lower (DAO) components, we adopted a two‐pronged approach. First, we performed gross anatomical dissections and collected morphometric measurements of muscle width and length, as described above. To address within‐individual non‐independence, bilateral samples from the same individual were averaged and represented as a single observation [63]. These data were visualized using the Uniform Manifold Approximation and Projection (UMAP), a dimensionality reduction technique [64]. To evaluate the discriminatory power of these data, we trained a support vector machine (SVM) classifier to distinguish human from NHP morphologies [65]. In parallel, we also applied principal component analysis (PCA) to the same dataset, as a complementary linear method for dimensionality reduction. In addition, to provide direct statistical support independent of dimensionality‐reduction methods, pairwise Wilcoxon rank‐sum tests were performed on the morphometric features. The threshold for statistical significance was set at p < 0.05. Histological Analysis Horizontal sections of the muscle were obtained at levels D1, D3, D5, D7, and D9 (Figure 4). To target the modiolar region, additional sections were taken at 15% of the muscle's total length above and below level D5. This sampling protocol yielded a total of 343 histological sections for human specimens (n = 49), and 154 sections for NHP specimens (n = 22). Following fixation in 4% paraformaldehyde for 48 h, samples were washed three times in phosphate‐buffered saline (PBS) and then processed in an automated tissue processor (Leica TP1020 Automatic Benchtop Tissue Processor). Processing included graded ethanol dehydration (70%, 90%, and 100%, with three changes in absolute ethanol), xylene clearing (three changes), and paraffin infiltration at 60°C. Total processing time ranged from 10.5 to 13 h, depending on specimen dimensions, with larger or thicker samples requiring longer processing to ensure adequate infiltration prior to microtome sectioning. Tissues were subsequently embedded in paraffin, sectioned at 5 µm thickness, and stained with hematoxylin and eosin for microscopic analysis. Images of all slides were obtained using the Olympus SlideViewer VS200. Fiber‐Level Quantification Rationale for Fiber‐Level Quantification To determine whether the CTC represents a single continuous muscle or two distinct muscular units (LAO and DAO), we quantified muscle fiber number at seven levels along the entire CTC and examined their spatial distribution relative to the labial commissure. As absolute fiber counts can vary with specimen dimensions, we normalized the counts using three complementary approaches: (1) number of fibers per unit cross‐sectional area (i.e., fiber density), (2) number of fibers per estimated total muscle volume, and (3) relative fiber proportion, defined by normalizing all levels to the specimen's maximal fiber count. In each analysis, we compared fiber counts between adjacent muscle levels to detect significant shifts in normalized values, which allowed us to assess whether the CTC is comprised of one or two muscles (LAO and DAO). This approach builds on the general architectural principle that fiber abundance peaks at the mid‐belly of a muscle and tapers toward its tendinous margin [66, 67, 68, 69, 70, 71, 72], enabling us to test whether the CTC conforms to a continuous pattern (i.e., shows no tapering toward the modiolus, consistent with the alternative hypothesis that the CTC is a single muscle) or instead exhibits a discontinuity at the labial commissure, as would be expected if two separate muscle bellies are present (the null hypothesis). Fiber‐Level Quantification Pipeline Fiber quantification was conducted using a semi‐automated full‐slide workflow that integrated deep learning‐based segmentation with manual validation. Whole‐slide histological images were subdivided into 2048 × 2048 pixel tiles for analysis. The mature model of MyoV [73], a deep‐learning algorithm optimized for muscle fiber detection, was used to segment and count individual fibers on each tile. Each analyzed slide was then manually reviewed in its entirety using the Count Tool in Adobe Photoshop (2025 version; Figure S2), and the automated output was corrected at the whole‐slide level so that fibers missed by the algorithm were manually added and any obvious over‐counting inaccuracies were resolved. No arbitrary subsampling of tiles or regions was used for validation. Thus, the final counts used for analysis represent corrected whole‐slide counts. Sections that were not quantifiable because of tissue tearing, staining failure, poor scanning focus, or other processing artifacts were not included in the analysis. In such cases, replacement sections were prepared from the same specimen at the relevant anatomical level, and only sections meeting the abovementioned quality criteria were entered into the final dataset. Muscle Fiber Count Standardization On the basis of fiber counts obtained at each level, three analyses were performed. First, muscle fiber density was calculated by dividing the number of fibers by the corresponding area on each slide. Tissue area was estimated using the Segment Anything Model [74], a deep learning tool for image segmentation. Second, total muscle volume was estimated by treating the muscle as a series of trapezoidal segments, each defined by the cross‐sectional areas at two adjacent levels and the distance between them. This approach allowed the approximation of the total muscle volume without assuming uniform geometry. Fiber counts at each level were then normalized by the total volume of the muscle to permit interspecific comparisons. Finally, the level exhibiting the highest fiber count within each specimen was identified, and all counts were normalized relative to this peak value. Statistical Analyses Muscle Fiber Count Distribution Along the Muscle All statistical analyses were performed using R (version 4.4.0). As the data were non‐normally distributed according to the Shapiro–Wilk test, log transformation was applied to normalize fiber count and fiber density distributions per species, followed by standardization (Figure S3). Statistical investigations of muscle fiber counts and peak locations along the muscle were conducted using the Mann–Whitney U test. To assess the presence of peaks in fiber distribution, peak detection analysis (see below) was applied at seven points, corresponding to the histological sampling sites described above (i.e., D1, D3, D5, D7, D9, and ±15% above and below D5). To compare the proportion of single versus multiple peaks along the muscle between humans and NHPs, 95% confidence intervals were calculated for both groups, and a chi‐squared test for equality of proportions was performed. Peak Detection Analysis Local peaks in muscle fiber number were identified for each specimen using a transition‐based algorithm. Transitions between adjacent sampling levels were classified as increasing (+1), stable (0), or decreasing (−1), on the basis of a threshold difference in fiber number. The threshold was chosen, based on the behavior of peak count results, to create a meaningful division peak count. A peak was defined as a transition from an increasing (+1) to a decreasing (−1) segment, representing a local maximum in fiber number. On the basis of the number of detected peaks, samples were categorized into two groups: specimens with a single peak and those with two or more peaks. To evaluate interspecific differences in peak distribution patterns, we calculated the proportion of specimens with multiple peaks, computed 95% confidence intervals for each group, and performed a chi‐squared test for equality of proportions. Classification Analysis We analyzed our dataset of muscle fiber counts using classification analysis, by applying dimensionality reduction methods, that is, UMAP, SVM, and PCA. Prior to analysis, transformations were applied to each feature within the dataset. If a feature followed a normal distribution, standard scaling was applied. For features following a log‐normal distribution, a natural logarithmic transformation was applied before standardization. If a feature did not conform to either a normal or log‐normal distribution, min–max normalization was applied. The UMAP algorithm was employed with the number of neighbors hyperparameter set to 15, preserving both global and local structure within the dataset. As UMAP is a nonlinear method that relies on the Euclidean distance between data points, applying equal standardization across all features ensured that each transformed feature had an equal influence on the final projection. The dataset was constructed at the individual level. When bilateral samples were available from the same individual, left and right muscle values were averaged and represented as a single observation. For classification, a linear SVM model with a regularization parameter C = 1 was applied. To determine the relative contribution of each feature (i.e., each level along the muscle) to the classification, the absolute value of each SVM weight was calculated and normalized by dividing by the total sum of the absolute weights [75]. To better visualize the relative contribution of discriminative features to sample separation, SVM‐derived feature importance values were applied as weights to the dataset prior to UMAP fitting, a process referred to as SVM‐weighted projection. This adjustment ensured that more discriminative features had a stronger influence on the final projection, improving cluster separability. As these projections were derived from the full dataset, they were used as descriptive visualizations to illustrate data structure and highlight the features contributing most strongly to group separation. Next, predictive classifier performance was assessed separately for each normalization strategy using repeated train/test evaluation. Specifically, in each iteration, the dataset was randomly divided into two equally sized groups (training and testing) using stratified random sampling to preserve the original class distribution, and this procedure was repeated using multiple splits (100 iterations). For each split, the classifier was trained on the training subset and evaluated on the held‐out testing subset. This allowed us to assess whether the trained model could accurately differentiate between human and NHP samples based on muscle fiber counts. Classifier performance was summarized using the median classification accuracy across repetitions, and feature‐importance stability was summarized as mean ± SD across runs. This procedure was used to provide a more robust estimate of classifier performance. Lastly, the same classification approach was applied to a set of morphometric measurements. These included the width ratios at different muscle levels (D1/D5 and D5/D9), as well as relative lengths of distinct muscle regions (L1/total muscle length and L2/total muscle length). These ratios were used to support interspecific morphological comparisons. Results Gross Anatomy In all the human specimens examined (n = 61), two distinct muscular bellies were identified. The LAO muscle originated from the canine fossa of the maxilla and converged toward the mouth corner, inserting into the modiolus. The DAO originated from the oblique line of the mandibular corpus as a fan‐shaped muscular sheet and converged into the modiolus as well. Figure 2 presents representative dissections showing the CTC muscles in situ. Figure 3 displays the harvested CTC muscles, aligned with the position of the labial commissure to facilitate anatomical orientation. In great apes (n = 4), the caninus muscle appeared as a single, continuous muscular sheet extending from the canine fossa of the maxilla to the level of the mandibular lower margin. The muscle consistently lay superficial to the platysma muscle and lacked any attachment to the mandible (Figures 2A–E and 3). In all other NHPs we examined, including Cercopithecidae (n = 20) and Platyrrhines (n = 2), a single muscular band originated from the maxillary canine fossa and descended toward the region of the labial commissure, where it typically terminated (Figures 2F–M and 3). Figure 5 summarizes the gross anatomical results. Comparative anatomy of the CTC across primates. Anterolateral views of the face in (A) human (H. sapiens); (B) chimpanzee (P. troglodytes); and (C) macaque (Macaca spp.). The skin is rendered partially transparent in the region surrounding the labial commissure to display the location and orientation of the CTC muscle fibers. See text for further explanation. Morphometric Measurements To evaluate shape differences in the CTC across species, we calculated four morphometric ratios derived from gross anatomical measurements: two width ratios (D1/D5 and D5/D9) and two length ratios (L1/total muscle length and L2/total muscle length) (see Section 2 and Figure 4). These ratios were chosen to capture variation in muscle tapering and vertical extent relative to the labial commissure. When projected into a low‐dimensional space using UMAP, human and other NHP specimens formed clearly separated clusters (Figure 6A). Human specimens clustered tightly, reflecting consistent proportions across individuals, whereas other NHPs grouped separately. Interestingly, all great ape specimens fell within or near the human cluster, suggesting a closer morphometric affinity to humans than to other NHPs. Gross morphometric analysis reveals clear separation of human, great ape, and nonhuman primate CTC architecture. Four morphometric ratios were calculated (see text): two width ratios (D1/D5 and D5/D9) and two length ratios (L1/total length and L2/total length). (A) UMAP projection of the dataset based on these standardized ratios showing preliminary clustering by species group. Each dot represents one individual specimen. (B) UMAP projection following feature importance reweighting by a support vector machine (SVM) classifier, which enhanced separation between groups. (C) Feature importance scores from the SVM model, indicating the relative contribution of each morphometric ratio to classification performance. Notably, the ratio D5/D9 (muscle width below the mouth corner) contributed 44.38%, more than any other feature. (D) Box plot of the D5/D9 ratio, the strongest contributor to group separation. NHPs exhibited the highest values, indicating relatively similar widths at D5 and D9, compared with both humans (p < 0.0001) and great apes (p = 0.0370). Great apes also showed higher values than humans (p = 0.0049). (E) Box plot of the D1/D5 ratio, showing higher values in humans than in great apes (p = 0.0057) and NHPs (p < 0.0001). (F) Box plot of the L1/total length ratio, representing the proportion of the muscle located above the labial commissure relative to the total CTC length. This ratio was highest in NHPs (80.55%), compared with humans (50.21%; p < 0.0001) and great apes (39.86%; p = 0.0087). Humans also showed higher values than great apes (p = 0.0171). (G) Box plot of the L2/total length ratio, which captures the proportion of the muscle that extends below the labial commissure. This single variable alone separates the three groups, with humans exhibiting significantly greater extension than nonhuman primates (p < 0.0001), and great apes falling between the two (see exact statistics in text). Boxes represent the interquartile range (IQR), center lines denote medians, whiskers extend to 1.5× IQR, and dots indicate individual specimens. This separation was reinforced by SVM classification (Figure 6B), in which all great apes were robustly assigned to the human cluster. Only one Cercopithecidae specimen (M. nigra) was misclassified as human, underscoring the model's overall discriminative power. Feature importance analysis identified the D5/D9 width ratio, representing muscle tapering below the labial commissure, as the most informative metric, contributing 44.38% to the classification decision. The D1/D5 ratio, reflecting upper tapering, contributed an additional 21.32% (Figure 6C). To further explore the vertical distribution of the muscle, we examined the L2/total length ratio, which quantifies the proportion of the CTC that extends below the level of the labial commissure (Figure 6G). Great apes exhibited the most pronounced continuation, with 60.14% of muscle length extending below the commissure. Humans showed slightly less continuation (49.79%), but still significantly more than other NHPs (19.45%). These differences were statistically significant across all group comparisons (great apes vs. humans: p = 0.0171; humans vs. other NHPs: p < 0.0001; great apes vs. other NHPs: p = 0.0087), highlighting the diagnostic value of this simple metric. A similar pattern was observed for the complementary L1/total length ratio, which was highest in other NHPs (80.55%) and lowest in great apes (39.86%), with humans showing intermediate values (50.21%); all pairwise comparisons were significant (p < 0.0171) (Figure 6F). The width ratios also distinguished the groups. D1/D5 was significantly greater in humans than in both other NHPs (p < 0.0001) and great apes (p = 0.0057), whereas great apes and other NHPs did not differ significantly (p = 0.9464) (Figure 6E). Conversely, D5/D9 ratio was lowest in humans (median = 0.2289), intermediate in great apes (median = 0.5333), and highest in other NHPs (median = 0.7659), with all pairwise comparisons reaching significance (human vs. other NHPs, p < 0.0001; human vs. great apes, p = 0.0049; great apes vs. other NHPs, p = 0.0370) (Figure 6D). Together, these results demonstrate that a small number of well‐chosen morphometric ratios, particularly the degree of muscle tapering and vertical extension below the mouth corner, are sufficient to robustly separate humans, great apes, and other primates based on CTC shape. Fiber‐Level Quantification Across the 497 sections analyzed in the present study, automated and manually corrected counts were strongly correlated (Spearman's ρ = 0.915). The median absolute correction was 674 fibers per section (IQR: 220–1580), corresponding to a median absolute percentage correction of 16.5% (IQR: 7.6%–36.0%). Manual corrections occurred in both directions, that is, undercounting or overcounting of the automated model, with 71 sections requiring downward adjustment of the initial automated count. Fiber Density Muscle fiber density was calculated for each level of the CTC by dividing the number of fibers by the cross‐sectional area. In humans, heatmap visualization of average fiber density across the D1–D9 levels revealed a localized reduction in labial commissure (level D5) (Figure 7A). Specifically, fiber density at D5 was significantly lower than at both levels D3 and D7 (adjusted p = 0.0226 for each comparison; adjusted p values were calculated using the Benjamini–Hochberg false discovery rate [FDR] procedure). In contrast, great apes and other NHPs exhibited relatively uniform density profiles along the vertical extent of the CTC, with no significant differences between these levels. Quantitative analysis of muscle fiber distribution along the vertical axis of the CTC. Heatmaps show the group‐level mean number of muscle fibers across sampling levels D1–D9 in humans (n = 49), great apes (n = 3), and other NHPs (n = 19). Fiber values are displayed as standardized log‐transformed intensities. Three normalization strategies were applied: (A) mean fibers density (fibers/mm2). (B) Fibers counts normalized to estimated total muscle volume. (C) Fiber counts normalized to maximal fibers value within each specimen. Within each group, Mann–Whitney U tests were used to compare group‐level means between: D3 versus D5 and D5 versus D7. In humans, panel (A) revealed significantly lower fibers density at the labial commissure (D5) compared to both D3 and D7 (adjusted p = 0.0226 for both comparisons). Also for humans, panels (B and C) show that fibers counts at D7 were significantly higher than at D5 (adjusted p = 6.85 × 10−12 and 6.77 × 10−16, respectively), consistent with increased color intensity. No significant differences were detected in great apes or other NHPs. Adjusted p values were calculated using the Benjamini–Hochberg false discovery rate correction. Peak detection analysis of fiber densities identified a bimodal distribution pattern in both humans and NHPs, with no significant difference in the proportions of the two peaks between groups (χ2 = 1.7158, d.f. = 1, p = 0.1902; Figure 8A). Among specimens with two peaks, the higher peak consistently occurred at D3 in both humans (n = 31) and other NHPs (n = 8), whereas in great apes, it shifted to D5 (n = 1) (Figure 8B). The lower of the two peaks was more variable, located between D5 and D7 in humans and other NHPs, and consistently at D7 in great apes. Notably, the ratio of the lower (distal) versus the higher (proximal) peak was significantly higher in humans compared to NHPs (p = 0.0227). Interspecific patterns of fiber peaks along the CTC. The number and anatomical location of fiber peaks were evaluated across three normalization strategies: (A–C) standardized log‐transformed fiber density; (D–F) fiber counts normalized to estimated muscle volume; (G–I) fiber counts normalized to specimen‐specific maxima. Left panels (A, D, G) display mean fiber distributions per group across levels D1–D9. Right panels depict the location of local peaks in individuals with either two peaks or a single peak: (B) two‐peak individuals based on fiber density (humans: n = 31; great apes: n = 1; other NHPs: n = 8); (C) single‐peak individuals based on fiber density (humans: n = 18; great apes: n = 2; other NHPs: n = 11); (E) two peaks in volume‐normalized data (humans: n = 29; great apes: n = 2; other NHPs: n = 4); (F) single peak in volume‐normalized data (humans: n = 20; great apes: n = 1; other NHPs: n = 15); (H) two peaks in maxima‐normalized data (humans: n = 27; great apes: n = 1; other NHPs: n = 4); (I) single peak in maxima‐normalized data (humans: n = 22; great apes: n = 2; other NHPs: n = 15). Species differences in the proportion of multi‐peak individuals were assessed using two‐tailed chi‐squared tests with continuity correction and 95% confidence intervals. Significant differences were found in panels (E) (χ2 = 6.52, d.f. = 1, p = 0.0107) and (H) (χ2 = 5.1, d.f. = 1, p = 0.0239), but not in panel (B) (χ2 = 1.72, d.f. = 1, p = 0.1902). Boxes represent the interquartile range (IQR), center lines denote medians, whiskers extend to 1.5× IQR, and dots indicate individual specimens. In specimens with a single density peak, the location of the peak differed across groups (Figure 8C). In other NHPs (n = 11), the peak was typically located between D3 and D5; in great apes (n = 2), it consistently occurred at D7; and in humans (n = 18), the peak location spanned the entire vertical axis of the muscle, indicating greater inter‐individual variability. A sensitivity analysis across a range of peak‐detection thresholds (0.14–0.26) yielded the same qualitative trend, with humans consistently showing a numerically higher proportion of specimens with at least two peaks than NHPs. However, the between‐group difference remained nonsignificant at the selected threshold (0.20; p = 0.1902) and at all neighboring thresholds tested, indicating that although the trend is stable, statistical support for a robust difference in fiber‐density peak number is limited (Figure S4A,B). Volume‐Normalized Fiber Counts Across Muscle Levels To account for interspecific variation in muscle size, we estimated the total CTC volume and normalized the fiber counts accordingly. In humans, a comparison of mean normalized fiber counts per volume revealed a statistically significant difference between levels D5 and D7, with higher values at D7 (adjusted p = 6.85 × 10−12; adjusted p values were calculated using the Benjamini–Hochberg FDR procedure), manifested also as an increase in color intensity in the heatmap (Figure 7B). No significant differences were observed in great apes or other NHPs. Peak detection analysis showed that the proportion of specimens exhibiting bimodal distribution was significantly higher in humans compared to other NHPs (χ2 = 6.5161, d.f. = 1, p = 0.01069; 95% confidence interval for the difference in proportions: 11.56%–64.71%) (Figure 8D). Among specimens with two peaks, the higher peak was typically located between D3 and D5, whereas the lower peak was found between D5 and D7 (n = 29 humans, n = 2 great apes, n = 4 other NHPs) (Figure 8E). In specimens with a single peak, its location varied by group: Other NHPs (n = 15) displayed a peak between D3 and above D5; in great apes (n = 1), it was consistently at D1; and in humans (n = 20), predominantly between D5 and below D7 (Figure 8F). To assess the sensitivity of this result to threshold choice, we repeated the peak‐detection analysis across a range of nearby thresholds (0.08–0.20). Across all tested values, humans consistently showed a higher proportion of specimens with at least two peaks than NHPs. At the selected threshold (0.14), this difference was statistically significant (p = 0.0107), and it remained significant at all tested thresholds (13/13), indicating that the observed between‐group difference is robust and not driven by a single threshold choice (Figure S4C,D). Normalized Fiber Counts Relative to Specimen‐Specific Maxima To further assess internal distribution patterns independent of absolute fiber counts or muscle size, we normalized fiber counts at each level (D1–D9) by the maximal fiber value within each specimen. Heatmap visualization revealed inter‐level variation across all groups (Figure 7C). In humans, fiber counts at level D7 were significantly higher than at D5 (adjusted p = 6.77 × 10−16; adjusted p values were calculated using the Benjamini–Hochberg FDR procedure), whereas no significant differences were detected between these levels in NHPs. Peak detection analysis showed that the proportion of specimens exhibiting bimodal patterns was significantly greater in humans than in NHPs (χ2 = 5.1, d.f. = 1, p = 0.0239; confidence interval for the difference in proportions: 7.38%–60.72%) (Figure 8G). Among bimodal specimens, the higher peak was typically located between D3 and D5, whereas the lower peak was found between D5 and D7, consistently across all groups (n = 27 humans, n = 1 great apes, n = 4 other NHPs) (Figure 8H). In specimens with a single peak, its location differed markedly across taxa: The peak was concentrated in the upper part of the CTC (D3 to above D5) in other NHPs (n = 15) and great apes (n = 2), whereas in humans, it was most frequently located in the lower part of the muscle, between D5 and D7 (n = 22) (Figure 8I). To evaluate the sensitivity of this result to threshold selection, we repeated the same peak‐detection pipeline across a range of nearby thresholds (0.20–0.32). Across all tested values, humans consistently showed a higher proportion of specimens with at least two peaks than other primates. At the selected threshold (0.26), this difference was statistically significant (p = 0.0239), and it remained significant for 8 of the 13 tested thresholds, indicating that the observed between‐group difference is not driven by a single threshold choice (Figure S4E,F). Classification Analysis UMAP projections of muscle fiber density across levels D1–D9 revealed partial separation between human and NHP specimens (Figure 9A). The application of the SVM classifier enhanced group separation, with great ape specimens positioned at the interface between the human and the other NHP clusters (Figure 9B). A feature importance analysis identified levels D3 and D7 as the most informative for classification, contributing 32.22% and 20.61%, respectively, to the model (Figure 9C). The model achieved a mean accuracy of 83.5% on the test set (Figure S5A–D). UMAP projection of muscle fiber distribution distinguishes humans from other primates. Panels are arranged by normalization strategy (columns): (A–D), fiber density; (E–H); fiber counts normalized to estimated muscle volume; (I–L), fiber counts normalized to specimen‐specific maxima. Within each normalization strategy, panels show the original UMAP projection, SVM feature importance, SVM feature importance, and a box plot of the most discriminative feature. (A, E, I) UMAP projections of the original dataset using all standardized features without weighting. Each point represents a single specimen. (B, F, J) UMAP projections after applying SVM‐based feature importance weighting. Features were scaled according to their relative importance, derived from a linear SVM classifier trained to distinguish humans from NHPs. Weighting enhanced group separation in the embedded space. (C, G, K) Tree maps displaying SVM feature importance scores. The area of each block corresponds to the relative contribution of a given muscle level (D1–D9) to the classification, calculated from normalized absolute SVM weights and expressed as a percentage. (D, H, L) Box plots of the most discriminative feature identified for each normalization strategy: D3 for fiber density (panel D), D7 for fiber counts normalized to estimated muscle volume (panel H), and D3 for fiber counts normalized to specimen‐specific maxima (panel L). Statistical comparisons are reported in the main text. Consistent with its importance in the classifier, fiber density at D3 was significantly higher in humans than in both great apes (p = 0.0267) and other NHPs (p < 0.0001), whereas no significant difference was detected between great apes and the other NHPs (Wilcoxon rank‐sum test; Figure 9D). When analyzing fiber counts normalized to the total muscle volume, UMAP projections did not reveal distinct clustering of human specimens, which appeared broadly distributed across the embedding space (Figure 9E). The application of the SVM classifier improved the group separation, with great apes clustering with the other NHPs (Figure 9F). A feature importance analysis identified D7 as the most discriminative level, contributing 40.23% to the model (Figure 9G). The classification achieved 91.7% accuracy on the test set (Figure S5E–H). Consistent with its importance in the classifier, D7 showed significantly higher values in humans than in both great apes (p = 0.0187) and other NHPs (p = 0.0002), whereas no significant difference was detected between great apes and the other NHPs (Wilcoxon rank‐sum test; Figure 9H). In the model based on fiber counts normalized to the maximal value per specimen, UMAP projections revealed a similar separation pattern (Figure 9I). The SVM classification achieved clear group separation, with great ape specimens clustering primarily with other NHPs (Figure 9J). A feature importance analysis identified D3 as the most informative level, contributing 24.80% to the model (Figure 9K). This model achieved 98.9% classification accuracy on the test set (Figure S5I–L). Consistent with its importance in the classifier, D3 showed significantly higher values in other NHPs (p < 0.0001) and great apes (p = 0.0187) than in humans, whereas no significant difference was detected between great apes and the other NHPs (Wilcoxon rank‐sum test; Figure 9L). To complement the UMAP and SVM results, we also applied PCA across all three datasets (Figure S6). The PCA yielded comparable group separation, with human specimens forming distinct clusters across all normalization strategies. This convergence across methods further supports the robustness of the observed anatomical differences in CTC fiber architecture. Discussion Anatomical Uniqueness of the DAO in Humans Our findings demonstrate that the DAO is an anatomically distinct muscle that is uniquely present in humans, within the primate sample examined in this study (mostly catarrhines). Through a multimodal approach, combining detailed dissection, morphometric analysis, and fiber‐level quantification, we show that NHPs, including great apes, lack a discrete mandibular muscle homologous to the human DAO. Detailed dissections across NHPs revealed a consistent pattern: All examined species possess a caninus muscle originating from the maxillary canine fossa and coursing inferiorly toward the labial commissure, where they interlace with adjacent perioral musculature (Figures 2, 3, 4). Although in some cases a few fibers extended slightly below the commissure, no specimen exhibited a distinct muscular belly anchored to the mandible. This configuration stands in marked contrast to the human condition, as well as to several published anatomical accounts of the NHP lower face [37, 38, 39, 40, 42, 43, 44, 45, 46, 47, 48]. Although the caninus muscle configuration described above was characteristic of most primate species examined, the few great‐ape specimens in our sample exhibited greater variation. In these cases, the muscular mass extended continuously over the modiolar region and reached the level of the mental foramen, as shown in Figure 5B, but notably, still without anchoring to the mandible. This further underscores the absence of a discrete mandibular depressor component analogous to that observed in humans. Our anatomical findings are corroborated by both morphometric and fiber‐level quantification analyses. Among the morphometric variables derived from gross dissection, the D5/D9 ratio, capturing the relative width of the muscle below the labial commissure, emerged as the most discriminative feature, contributing 44.38% to SVM‐based group classification. This single metric alone was sufficient to separate humans, great apes, and other primates in a structured and anatomically interpretable manner. Specifically, it reflects the pronounced triangular tapering of the human DAO, which narrows toward its insertion in the mouth corner. In contrast, NHPs exhibit a more uniform‐width configuration, consistent with the traditional description of the CTC as a continuous sheet [26, 27]. These results support the notion that a distinct DAO with a triangular profile is unique to the human face, reinforcing the anatomical dissociation between human and NHP lower facial musculature. A plausible explanation for why the absence of a discrete DAO in NHPs has gone largely unrecognized in prior anatomical accounts relates to the complex layered organization of the superior and lateral perioral musculature. In primates, the quadratus labii superioris forms a broad, sheet‐like layer that covers the superior labial region and effectively veils the bulk of the caninus muscle, which lies deep to it and courses obliquely toward the labial commissure [28, 29, 30, 32, 34]. Depending on variation, a small subset of caninus fibers may traverse the inferior margin of the quadratus labii superioris and extend below the commissure, creating the impression of a lower lip‐depressing component. We propose that this limited and variably exposed fiber segment has frequently been interpreted as a homolog of the human DAO in earlier studies. However, our dissections demonstrate that these fibers consistently lack a mandibular attachment and instead terminate within the modiolar or perioral connective tissue complex, rendering them unlikely to generate DAO‐mediated human‐like mouth‐corner depression. By integrating morphometric and fiber‐level analyses, the present study resolves this ambiguity and clarifies that the apparent DAO‐like structure in NHPs represents a superficial extension of the caninus rather than a distinct mandibular depressor muscle homologous to the human DAO. Quantitative Evidence for DAO Uniqueness Although raw fiber density profiles did not reveal major differences in peak distribution between humans and NHPs, normalization of fiber counts to either total muscle volume or to maximal fiber number uncovered several key distinctions. NHPs typically show a single peak in fiber number above the labial commissure, corresponding to the caninus muscle, and homologous to the human LAO muscle. Fiber counts then decline toward the lower margin, consistent with the myotendinous junction [66, 69, 76]. By contrast, human samples exhibit a bimodal pattern, with peaks both above and below the commissure, consistent with two anatomically distinct bellies: the LAO and DAO, respectively. These structural differences were further reinforced by machine learning classification. When fiber counts were normalized to total muscle volume, unsupervised UMAP projections and SVM analyses achieved 91.7% accuracy in distinguishing humans from NHPs, with level D7 contributing 40.30% ± 0.77% to the classification (Figure 9E–H). When normalized to the specimen's maximal fiber value, classification accuracy increased to 98.9%, with D3 contributing 23.18% ± 0.40% to the model (Figure 9I–L). Together, these findings reinforce the conclusion that the DAO is absent in the lower face of NHPs and support its human specificity. Importantly, great ape specimens, although few, considerably increase the range of interspecific variation captured in the present study. Their intermediate position in morphometric and fiber‐level space should not, however, be taken to imply homology or near‐equivalence with the human condition. Rather, it reflects the fact that in both great apes and humans the CTC spans the full distance between the canine fossa and the mandible, crossing the modiolar region. This broad architectural similarity influences clustering based on overall morphometric configuration. Crucially, however, similarity in such broad features does not entail the presence of the mechanically and architecturally critical feature of interest: a discrete mandibular DAO belly. In humans, the CTC is subdivided at the mouth angle by a musculotendinous node, and its inferior belly is distinctly anchored to the mandible and the chin. In great apes, by contrast, the muscle mass may extend across the modiolar region but remains continuous, lacking both a separable inferior belly and a discrete mandibular attachment. Thus, the intermediate clustering of great apes likely reflects shared aspects of overall facial architecture and mass distribution, rather than a true precursor of the human configuration. AU15 and the Expressive Repertoire of the Primate Face The DAO draws the mouth corners distinctly downward [20], providing the anatomical basis for AU15—the lip corner depressor—as defined in the FACS [52]. In humans, this action has long been associated with the facial expression of sadness, although it is not restricted to sadness‐related displays and may also occur in configurations associated with disgust, pouting, and other negative affective states [20, 52, 77, 78]. More generally, in both humans and NHPs, a single muscle movement may participate in several distinct facial displays [4]. It is therefore worth considering whether AU15 may be recruited into facial signals beyond sadness in humans, and how such signals may resemble or differ from those of other primates. Chimpanzees, for example, produce vocalizations strongly reminiscent of crying (whimpering) [79], yet these are accompanied by a lip configuration more similar to the bared‐teeth scream face. As noted by Preuschoft [11], crying in chimpanzees is commonly associated with a stretched pout–whimper, in which the teeth are partially bared and the mouth corners are horizontally retracted, a configuration not typically seen in humans. Likewise, in the chimpanzee frustration‐sadness expression, the lower lip is extended and pulled downward, whereas in macaques, the pout face involves closed or only slightly open jaws, with the lips contracted vertically and horizontally into a pursed configuration [16]. Taken together, these observations suggest that, unlike the human condition, AU15 does not appear to contribute to these facial signals in other primates in the same manner. Our findings suggest that NHPs are unlikely to generate DAO‐mediated mouth‐corner depression due to the absence of the requisite musculature. Although this interpretation challenges the prevailing view that NHPs possess a DAO homolog (triangularis muscle), it aligns with cumulative findings from cross‐species adaptations of the FACS. Notably, AU15 has never been documented in any validated NHP FACS adaptation to date [54, 55, 56, 57, 58, 59, 60, 61]. A single exception is found in Waller et al. [80], who reported an AU15‐like movement in Macaca mulatta during electromyographic stimulation of the face (documented in Supporting Information 11 of that study). However, we suggest that this movement is more consistent with activation of the platysma than with a discrete DAO homolog. In many NHPs, the platysma extends broadly across the lower face and may reach the modiolar and lower lip region [28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 81]. Its contraction can therefore influence commissural position and generate a broader lower face retraction under stimulation, corresponding to AU20 and/or AU21. As described by Ekman and Friesen [82], AU20/21 are characterized primarily by lateral or caudolateral stretching and retraction of the lips and adjacent soft tissues, often with visible tension in the neck, whereas AU15 is defined by a more localized downward displacement of the lip corners. Importantly, such a platysma‐mediated retraction differs from the focal depression of the labial commissure produced by the discrete mandibular DAO in humans. Thus, although stimulation of the platysma may affect the position of the mouth corner, the resulting movement lacks the localized precision and affective configuration of AU15 as it occurs in the human face. The premise that “it is unlikely that the down‐turned mouth component has continuity across species as this action unit (AU15) does not appear in any FACS outside of humans” [p. 10] [62] lends support to our interpretation. Uniqueness of the DAO in Humans and a Novel Hypothesis of Helplessness Signaling Importantly, the principal contribution of the present study is anatomical, as it provides only anatomical evidence for the unique architecture of the human DAO within the primate sample examined here. At the same time, these findings allow a cautious consideration of the evolutionary context in which such a muscle may have arisen, and any functional interpretation offered here should therefore be regarded as a testable hypothesis rather than as a direct implication of the data. One possible framework is the distinctive human life history. Compared to other primates, humans are characterized by highly altricial neonates, prolonged postnatal brain growth, delayed maturation, and an extended childhood between weaning and nutritional independence [83, 84, 85, 86, 87, 88]. This extended developmental trajectory has been widely interpreted as an adaptive feature that supports rapid brain growth and social learning within a culturally enriched environment [86]. It has also been hypothesized that secondary altriciality evolved in Homo erectus [89, 90, 91, 92], highlighting the significance of prolonged childhood [86, 93, 94, 95, 96, 97, 98]. A pivotal shift in hominin life history was the loss of infant clinging, a likely consequence of bipedal locomotion [99, 100] and the loss of maternal body hair [88], which would have reduced continuous physical contact between mothers and offspring [101]. This constraint may have necessitated the evolution of enhanced visual signaling to maintain caregiver proximity [102]. As noted by Hrdy and Burkart [103], among great apes, “newborns constantly held by mothers in direct skin‐to‐skin contact have less need to smile or vocalize” (p. 4). Over evolutionary time, distal mother–infant gestural communication [104] and affective vocalizations [6, 23, 105] likely became more frequent and salient to compensate for the reduction in tactile cues. Within this evolving communicative landscape, the emergence of a novel expressive movement may have conferred a selective advantage by amplifying infant visibility and more effectively signaling distress. This interpretation is supported by Hrdy's [88] affirmation that “the more direct physical contact there is between a mother and her infant, the less time each party spends looking into the face of the other” (p. 120), further underscoring the evolutionary significance of facial signaling in the absence of sustained bodily contact. Human infants, unlike other apes, appear to have been under greater evolutionary pressure to solicit attention not only from their mothers but also from allomothers [106]. This hypothesis may also be situated within the broader context of human caregiving ecology. In cooperative childcare systems, human infant survival likely depends not only on prolonged postnatal care, but also on the ability to recruit attention and responsiveness from multiple caregivers, including mothers, allomothers, and other alloparents [107, 108, 109]. Under such conditions, increasingly legible distal social‐affective signals may have become especially valuable, particularly as reduced continuous mother–infant bodily contact placed greater weight on visual and vocal communication [88, 103]. Within this framework, we hypothesize that the unique human DAO may be viewed as an anatomical specialization whose expressive effect could have increased the visual salience of infant distress and thereby facilitated the recruitment of caregiving responses. Importantly, as highlighted by several authors, the pre‐cry face (e.g., pout and “kidney mouth”), often directed at the caregiver, can serve as early warning signals, which allow the caregiver to act and avert a full‐blown distress [21, 110, 111]. We highlight that this possibility is presented as a plausible co‐evolutionary context for the emergence of the DAO, rather than as a direct conclusion of the anatomical data. Furthermore, AU15 participates in the infant and child cry, generally interpreted as an honest signal of need effective in soliciting caregiver attention [112, 113]. Of particular relevance, pouting is distinguished from the cry face, although AU15 participates in both. According to Oster [21], infant facial expressions should be examined as ontogenetic adaptations, that is, biologically based adaptations crucial for the infant's survival. Specifically, regarding the facial muscles involved in pouting and its temporal association with crying, she argued [78, 114] that this expression reflects an effort to regulate distress, and that it represents “a salient signal to caregivers that the infant is upset (for whatever reason) and is likely to cry if the situation is not improved” (p. 284) [21]. Along these lines, the distinction between high‐intensity vocal crying and the silent, low‐intensity pout that may precede escalation is also significant. The pout may function as a graded, lower cost signal that fine‐tunes caregiver response before distress escalates, especially given that crying itself carries an energetic cost [23]. Also supporting this idea, Huebner and Izard [115] suggested that infant facial expressions of distress may be a more discriminative component of the infant's message than the cry itself. This view further suggests that DAO‐mediated signaling may participate in a broader communicative and regulatory landscape than crying alone. Beyond the broader evolutionary interpretation proposed here, the anatomical findings of the present study also generate a set of testable predictions that may help ground this hypothesis empirically. Developmentally, if the DAO is linked to signaling during early dependency, its distinctive architecture would be expected to emerge early in human ontogeny and to be most functionally salient during infancy and early childhood, with the associated movement potentially changing in frequency, form, or communicative role across the lifespan. Behaviorally, AU15‐like or other homologous expressions would be expected to occur more frequently, and in more specific caregiving contexts, in human infants than in nonhuman ape infants. In this respect, recent work has emphasized that comparative research should move beyond static, prototypical facial displays toward the study of dynamic facial behavior in naturalistic social interaction, while emerging automated cross‐species tools now offer increasingly fine‐grained ways to track such movements and to clarify what, if anything, is truly unique to the human face [116, 117]. Mechanistically, facial displays involving AU15 may be predicted to play a significant role in modulating caregiver attention, responsiveness, proximity, or intervention, particularly in pre‐cry or low‐intensity distress contexts. Overall, this study provides anatomical evidence for the uniqueness of the human DAO and suggests that the absence of a comparable muscular configuration in other primates may help explain why a human‐like AU15 has not been identified in other primate FACS adaptations. Building on this anatomical finding, we propose a new hypothesis, interpreting the DAO as a signal for helplessness in human infancy and prolonged childhood. To evaluate this evolutionary hypothesis, further research is required, including fossil analysis and behavioral observations on caregiver–infant interaction in humans and other primates, to trace the emergence of this trait and its relevance to the evolution of human life history. Limitations Two important limitations of the present study should be acknowledged. First, the great ape sample was necessarily small, owing to the ethical and practical difficulty of obtaining such specimens, and individual variation within great apes may therefore not be fully captured here. Second, although the present findings support the conclusion that a discrete mandibular DAO is unique to humans within the primate sample examined in this study (mostly catarrhines), further comparative work across a broader range of primate taxa will be necessary to assess the full generality of this pattern across the primate order. Finally, although the machine‐learning analyses yielded patterns consistent with the anatomical, histological, and morphometric results, their interpretation should remain proportional to the available sample sizes, and they are therefore best viewed as complementary to the principal anatomical evidence presented here. Author Contributions Liat Rotenstreich and Assaf Marom conceived and designed the study. Liat Rotenstreich, Assaf Marom, Nir Edery, and Lior Moss procured the human and nonhuman primate craniofacial specimens. Liat Rotenstreich and Assaf Marom carried out the dissections and microanatomical analyses. Aya Vituri, Moni Shahar, Liat Rotenstreich, and Assaf Marom performed the statistical analyses. Kfir Eliyahu, Ori Farkas, Liat Rotenstreich, and Assaf Marom performed the machine‐learning‐based analyses. Liat Rotenstreich and Assaf Marom wrote the manuscript. All authors revised and approved the final manuscript. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Supplementary Figure: nyas70291‐sup‐0001‐FigureS1.tif Supplementary Figure: nyas70291‐sup‐0002‐FigureS2.tif Supplementary Figure: nyas70291‐sup‐0003‐FigureS3.tif Supplementary Figure: nyas70291‐sup‐0004‐FigureS4.tif Supplementary Figure: nyas70291‐sup‐0005‐FigureS5.tif Supplementary Figure: nyas70291‐sup‐0006‐FigureS6.tif Supplementary File: nyas70291‐sup‐0007‐SuppMat.docx