Gross anatomical dissections in the scapular, humeral joint, and brachial regions were conducted on five P. flavus specimens, which were previously employed in a study of the extrinsic thoracic limb muscles (Vélez-García and Miglino 2023). Both thoracic limbs were dissected and meticulously identified (Table 1). The anatomical characteristics were described following the terminology of the Nomina Anatomica Veterinaria (International Committee on Veterinary Gross Anatomical Nomenclature 2017). However, additional terms were incorporated due to the presence of other muscles and heads, such as the caput breve and caput longum of the m. biceps brachii, m. coracobrachialis brevis, m. coracobrachialis longus (Beswick-Perrin 1871; Windle and Parsons 1897; Julitz 1909), m. anconeus lateralis, and m. anconeus medialis (Barone 2020; Enciso-García and Vélez-García 2022). Photographs of the dissections were captured using a Canon T5i camera paired with a 60 mm macro lens and an EOS 6D camera paired with a 100 mm macro lens. This investigation received ethical approval from the bioethics committee of the Universidad del Tolima (2.3-059).

The comparative analysis of the intrinsic shoulder and brachial muscles of caniforms can be framed as a classification problem of categorical variables or a clustering problem. Clustering represents an unsupervised machine learning technique utilized to identify and segregate natural groups based on the inherited properties within a dataset (Müller and Guido 2017; Kotu and Deshpande 2019). The DBSCAN (Density-Based Spatial Clustering of Application with Noise) algorithm, following a nearest neighbor distance approach, was designed to identify clusters with varying shapes and outliers (noise) (Ester et al. 1996). Considering the literature review (Table 2; see the discussion section), the muscles exhibiting the most variability among caniforms are the teres minor, coracobrachialis brevis, coracobrachialis longus, biceps brachii, triceps brachii, tensor fasciae antebrachii, and anconeus medialis. Consequently, these muscles were designated as the main categorical variables (i.e., Tmin, CBb, CBl, BB, TB, TFA, AM, respectively). The study engaged DBSCAN to cluster the studied species based on these seven variables, and its predictive performance was measured using the silhouette coefficient (Rousseeuw 1987). The DBSCAN was applied using Scikit-learn 1.2.2 in Jupyter Notebook 6.5.2 (Kluyver et al. 2016). The final cluster visualization was generated using Tableau 2021.4.

For each specimen, the seven categorical variables underwent an encoding process to assign an integer value class corresponding to the muscle variants. Teres minor (Tmin), coracobrachialis brevis (CBb), coracobrachialis longus (CBl) and anconeus medialis (AM) variables were encoded into two classes, denoted as absent (0) or present (1). “Absent” was indicated when the muscle was either not described or reported as vestigial or fused with other muscle (e.g., m. teres minor was considered absent when reported as fused with m. infraspinatus in P. flavus as described by Beswick-Perrin 1871; Windle and Parsons 1897; Julitz 1909). In some cases, the muscle was classified as “present” (1) when an author did not explicitly mention it, but other authors had reported its presence, and it was consistent with the family’s tendency to possess it. For example, m. anconeus medialis is considered present in the P. lotor specimens of Feeney (1999), since other authors had reported its presence in this species (Allen 1882; Windle and Parsons 1897). In addition, this muscle aligns with the typical characteristics of the Procyonidae family.

The variables triceps brachii (TB) and biceps brachii (BB) were numerically encoded into four classes based on the number of heads (capita). Therefore, values ranging from two (2) to five (5) were considered for each muscle. The reported caput angulare of the m. triceps brachii in mustelids was regarded as the caudal part of the m. tensor fasciae antebrachii (see the discussion section).

Finally, the tensor fasciae antebrachii (TFA) variable was encoded into five classes. “Absent” (0) was designated only when the author explicitly reported its complete absence (e.g., Böhmer et al. 2020 reported the absence of this muscle in Vulpes vulpes and Cuon alpinus). The presence of the cranial portion (1) was indicated when the muscle was solely reported to originate from the common tendon of the latissimus dorsi and teres major muscles (e.g., most canids). The presence of the caudal portion (2) was indicated when it originated from the m. teres major, caudal angle, or caudal margin of the scapula. The presence of both the cranial and caudal portions (3) was recorded (e.g., P. flavus). The presence of three portions (4) was considered in Ursus americanus based on Shepherd’s (1883) descriptions. This last characteristic was also extended to Ursus maritimus, as Kelley (1888) compared his findings with those of Shepherd (1883) in U. americanus.

It is important to note that some specimens were excluded from the statistical analysis because only one of the two anatomical regions was studied by the authors, either the shoulder or the brachium (e.g., Davis 1949; Pereira et al. 2010). However, anatomical description from U. americanus were considered because a subsequent publication concerning Ailuropoda melanoleuca highlighted differences between U. americanus and Tremarctos ornatus (Davis 1964).

DBSCAN identified five groups as the optimal number of clusters of specimens based on the statistical analysis of the absence and presence of intrinsic shoulder and brachial muscles. The DBSCAN performance was considered good, with a silhouette coefficient of 0.65. It is worth noting that a minimum value of 0.5 is generally accepted as indicative of good clustering. Most P. flavus specimens (including those referenced in Beswick-Perrin 1871; Julitz 1909; Böhmer et al. 2020) were assigned to the third cluster. This cluster also included most specimens of the ailurid Ailurus fulgens and the ursids of the genera Ursus and Tremarctos. The defining characteristics of this group is the presence of most muscles. In the first cluster, one P. flavus limb (PfS4-F4R) is found, primarly due to the absence of the caput breve of the m. biceps brachii. This cluster also included mustelids of the genera Martes, Pekannia, and Eira, two A. fulgens specimens, and the ursids Ailuropoda melanoleuca and Melursus ursinus. Two P. flavus specimens (PfS2-M2 and one from Windle and Parsons 1897) formed the fourth cluster, along with the other genera of procyonids and the mustelids Galictis cuja and Meles meles. The primary characteristics of this group are the absence of the m. coracobrachialis longus and the caput breve of the m. biceps brachii. Most canids clustered in the second group due to the absence of m. coracobrachialis longus, m. anconeus medialis, and the caudal part of the m. tensor fasciae antebrachii. However, Cuon alpinus was an exception and located within the fourth cluster due to the presence of m. anconeus medialis. Lutrines (mustelids) and mephitids clustered in the fifth group due to the absence of the teres minor, coracobrachialis brevis, and coracobrachialis longus muscles. The presence of the mustelid Taxidea taxus in this cluster was due to the absence of both mm. coracobrachiales. Additionally, one Aonyx sp. limb (Lutrine) was located in the fourth cluster due to the presence of the m. coracobrachialis brevis. Figure 10 provides a visual representation of the above-mentioned clusters for reference.

The origin and insertion formerly reported for the supraspinatus and infraspinatus muscles in P. flavus were similar to the specimens of the present study but the origin of both muscles from the acromion (Julitz 1909). This latter origin is also present in other procyonids, such as Nasua nasua, Procyon cancrivorus (Tarquini et al. 2023) and Procyon lotor (Feeney 1999). Several authors generally describe minimal differences in the supraspinatus and infraspinatus muscles among procyonids (Beswick-Perrin 1871; Mackintosh 1875; Allen 1882; Windle 1888; Windle and Parsons 1897; Beddard 1900). However, it is worth noting that some procyonids, including P. lotor (Allen 1882), P. cancrivorus (Santos et al. 2010b) and N. nasua (Santos et al. 2010a; Tarquini et al. 2023) may exhibit origins from an intermuscular septum cranially to the cranial margin of the scapula. In P. cancrivorus, the supraspinatus insertion extends to the transverse humeral retinaculum (Tarquini et al. 2019) and may exhibit variations with two bellies in P. cancrivorus (Tarquini et al. 2023; Vélez García et al. 2023), and P. lotor (Allen 1882). This is distinct from most non-procyonid caniforms, which typically possess one belly as seen in P. flavus, while it normally has two bellies in canids (Feeney 1999; Vélez-García et al. 2018b) and three bellies in the mustelid G. cuja (Ercoli et al. 2015).

In the present study, the m. teres minor was completely separated from the m. infraspinatus in all P. flavus specimens, consistent with reports in the same species (Böhmer et al. 2020) and other procyonids (Davis 1949; Feeney 1999; Böhmer et al. 2020; Tarquini et al. 2023). However, it differs from previous descriptions, where fusion with the m. infraspinatus was documented in P. flavus (Beswick-Perrin 1871; Windle and Parsons 1897; Julitz 1909) and occasionally in P. lotor (Allen 1882). Origins from the infraglenoid tubercle, as found in P. cancrivorus (Santos et al. 2010b) and N. nasua (Santos et al. 2010a), were not observed in P. flavus . Among most non-procyonid caniforms, the m. teres minor is typically separated from the m. infraspinatus. However, fusion with the m. infraspinatus has been reported in the mustelid Enhydra lutris (Howard 1973), the mephitid Mephitis mephitis (Hall 1926), and the ursid Ursus americanus (Shepherd 1883). In some species, the muscle may be vestigial or entirely absent, as seen in the mephitid Spilogale gracilis (Hall 1926), the mustelids Aonyx sp. (Macalister 1873a) and Lutra lutra (Haughton 1864c), and occasional cases in G. cuja (Ercoli et al. 2015).

While examining the m. subscapularis, it was found that its attachments do not differ among procyonids, corroborating previous descriptions (Beswick-Perrin 1871; Allen 1882; Windle 1888; Julitz 1909; Santos et al. 2010a, 2010b). Nevertheless, the number of bellies varies across species, with P. lotor showing two to four bellies (Allen 1882; Windle and Parsons 1897; Davis 1949; Feeney 1999), N. narica exhibiting three, N. nasua presenting six (Mackintosh 1875), and P. flavus displaying up to eight. Additionally, the muscle may insert via two tendons onto the lesser tubercle in species as Bassariscus astutus, P. lotor, N. narica (Davis 1949), and N. nasua (Tarquini et al. 2023). In non-procyonid caniforms, the muscle has been reported to have a maximum of six bellies in G. cuja (Ercoli et al. 2015), Martes martes (Yousefi et al. 2018), Canis lupus familiaris (Hermanson 2020), and Cerdocyon thous (Vélez-García et al. 2018b). The insertion via two tendons has only been reported in U. americanus (Davis 1949) and M. ursinus (Annie et al. 2017).

The m. deltoideus of P. flavus closely resembled descriptions provided by other authors (Beswick-Perrin 1871; Windle and Parsons 1897; Julitz 1909; Böhmer et al. 2020) and other procyonids (Windle 1888; Windle and Parsons 1897; Santos et al. 2010a, 2010b; Tarquini et al. 2023). However, the scapular part of the muscle may also originate from the infraspinatus fascia in N. nasua (Mackintosh 1875; Tarquini et al. 2023) and as a variation in P. lotor (Allen 1882). Additionally, the tendon has contact with the caput laterale of the m. triceps brachii in P. lotor (Allen 1882), while in N. narica, the insertion extends distally onto the humerus (Mackintosh 1875). The acromial part of the m. deltoideus may have some continuous fibers with the m. brachialis or inserts onto this muscle in N. nasua and N. narica (Mackintosh 1875; Tarquini et al. 2023). The division of the m. deltoideus into two parts is generally present in all caniforms, with each part inserting separately onto the deltoid tuberosity in G. cuja (Ercoli et al. 2015), Aonyx sp. (Macalister 1873a), A. fulgens (Fisher et al. 2009), and M. ursinus (Annie et al. 2017). It is important to highlight that the clavicular part (pars clavicularis) of the m. deltoid, mentioned in previous studies (Beswick-Perrin 1871; Mackintosh 1875; Allen 1882; Windle 1888; Windle and Parsons 1897; Julitz 1909), is designated as the m. cleidobrachialis (International Committee on Veterinary Gross Anatomical Nomenclature 2017), since it originates from the clavicular intersection. Therefore, it is an extrinsic thoracic limb muscle because the clavicle is an absent or vestigial bone in procyonids (Souza Junior et al. 2020; Vélez-García and Miglino 2023). Nonetheless, in procyonids, this muscle has evolved from the m. deltoideus and retains innervation by the axillary nerve (Enciso-García and Vélez-García 2022; Vélez García et al. 2023; Vélez-García and Miglino 2023).

The m. teres major in P. flavus originates from the proximal third of the caudal margin of the scapula, consistent with previous descriptions (Windle and Parsons 1897; Julitz 1909). However, this muscle also originates from the caudal angle and the m. subscapularis, a characteristic not reported by other authors (Beswick-Perrin 1871; Windle and Parsons 1897). A similar origin is also found in P. lotor (Allen 1882; Feeney 1999), P. cancrivorus, and N. nasua (Santos et al. 2010a, 2010b; Tarquini et al. 2023). Additionally, an origin from the m. infraspinatus has been observed in N. nasua (Tarquini et al. 2023) and as a variation in P. lotor (Allen 1882). Among non-procyonid caniforms, the m. teres major exhibits considerable variation in its origin, while the insertion consistently occurs via a common tendon with the m. latissimus dorsi.

The arrangement of the coracobrachialis brevis and longus muscles in P. flavus are consistent with previous descriptions (Beswick-Perrin 1871; Julitz 1909; Davis 1964). The m. coracobrachialis longus may be absent in P. flavus (Windle and Parsons 1897), as similarly observed in one specimen of the present study. Böhmer et al. (2020) only described one muscle for P. flavus, although both muscles can be observed in the figures of their study. Therefore, based on the total limbs reported by several authors for P. flavus (Beswick-Perrin 1871; Windle and Parsons 1897; Julitz 1909; Davis 1964; Böhmer et al. 2020; present study), the prevalence of the absence of the m. coracobrachialis brevis would be 20% (4/20 limbs). In contrast, other procyonids typically have only one m. coracobrachialis (Mackintosh 1875; Allen 1882; Beddard 1900; Davis 1949; Santos et al. 2010a, 2010b; Böhmer et al. 2020; Tarquini et al. 2023), which, based on its insertion, corresponds to the m. coracobrachialis brevis. However, both muscles have been assigned with different names in the literature (Table 3). The insertion of the m. coracobrachialis brevis onto the intertubercular groove, as reported by Julitz (1909), was not observed in our study. The insertion of the m. coracobrachialis longus proximal to the supracondylar foramen was also reported in one study (Beswick-Perrin 1871), while another study reported it onto the medial supracondylar crest (Julitz 1909). In non-procyonid caniforms, both muscles may be present only in the ailurid A. fulgens (Carlsson 1925; Fisher et al. 2009), all ursids (Shepherd 1883; Kelley 1888; Windle and Parsons 1897; Davis 1949, 1964; Annie et al. 2017), and some mustelids, with a variant configuration. Both muscles are present in Pekania pennanti (Leach 1977; Feeney 1999), Martes americana (Leach 1977), M. foina (Mackintosh 1875), M. martes (Yousefi et al. 2018), and Martes caurina (Hall 1926). The m. coracobrachialis longus is uniquely present in E. barbara (Macalister 1873b), M. martes (Böhmer et al. 2018), and M. foina (Böhmer et al. 2018). Böhmer et al. (2018) initially described the insertion of this muscle (“m. epitrochlearis”) onto the olecranon. However, in a subsequent study, the insertion was described as occurring on the middle third of the humerus (Böhmer et al. 2020), which is consistent with other authors reporting in M. caurina (Hall 1926) and M. martes (Yousefi et al. 2018). In other mustelids, the m. coracobrachialis longus inserts proximal to the supracondylar foramen, similar to P. flavus (Macalister 1873b; Mackintosh 1875; Leach 1977; Feeney 1999). In G. cuja, only the m. coracobrachialis brevis is typically present, although it might be absent (Ercoli et al. 2015). In other mustelids, both coracobrachialis muscles are entirely absent, as observed in Aonyx (Macalister 1873a), L. lutra (Haughton 1864), E. lutris (Howard 1973), and T. taxus (Moore et al. 2013). In mephitids, both coracobrachiales muscles are also totally absent in M. mephitis and S. gracilis (Hall 1926). In the ailurid A. fulgens and all ursids, the m. coracobrachialis longus reaches the medial epicondyle region (Shepherd 1883; Kelley 1888; Windle and Parsons 1897; Carlsson 1925; Davis 1949, 1964; Fisher et al. 2009; Annie et al. 2017).

In this study, we found that the m. biceps brachii in P. flavus exhibited two heads, as that reported previously (Beswick-Perrin 1871; Windle and Parsons 1897; Julitz 1909; Davis 1964; Böhmer et al. 2020). However, it is important to note that none of these prior studies reported the absence of the caput breve as an anatomical variant. In contrast, other procyonids typically have only one caput, corresponding to the caput longum, which inserts onto the radius (Allen 1882; Windle 1888; Windle and Parsons 1897; Beddard 1900; Davis 1949; Böhmer et al. 2020; Tarquini et al. 2023). Notably, the caput breve exhibited a feebly shape in P. lotor (Windle and Parsons 1897), and additional insertions onto the ulna were reported in P. lotor (Feeney 1999), P. cancrivorus (Pereira et al. 2010; Santos et al. 2010b) and N. nasua (Santos et al. 2010a). In non-procyonid caniforms, the presence of the caput breve was only found in the ailurid A. fulgens (Carlsson 1925; Davis 1964; Fisher et al. 2009) and ursids of the genera Ursus and Tremarctos (Shepherd 1883; Kelley 1888; Windle and Parsons 1897; Davis 1949, 1964). However, it was absent in the ursids A. melanoleuca (Davis 1964) and M. ursinus (Annie et al. 2019), and occasionally in A. fulgens (Fisher et al. 2009). Based on the reports from several studies, in A. fulgens (Carlsson 1925; Davis 1964; Fisher et al. 2009) and P. flavus (Beswick-Perrin 1871; Windle and Parsons 1897; Julitz 1909; Davis 1964; Böhmer et al. 2020; present study), the prevalence of the absence of the caput breve was 33% (4/12 limbs) and 10% (2/20), respectively. In other caniforms, the absence of the caput breve of m. biceps brachii is the common pattern.

Our findings suggest variations regarding the origin of the m. brachialis, particularly in P. flavus. While previous authors (Beswick-Perrin 1871; Windle and Parsons 1897; Julitz 1909) did not report the lateral supracondylar crest as an origin in P. flavus, our study identified this as an additional origin. Other origins have been reported from the proximal half of the humerus (Beswick-Perrin 1871), and along the lateral surface of the humerus and humeral crest (Julitz 1909). In other procyonids, the origin and insertion patterns are similar to P. flavus (Mackintosh 1875; Allen 1882; Feeney 1999; Tarquini et al. 2023). However, in P. cancrivorus and N. nasua (Pereira et al. 2010; Santos et al. 2010a, 2010b), the muscle may only originate from the proximal part of the lateral surface of the humerus. Further, it may even insert onto the radial and ulnar tuberosities in both species (Santos et al. 2010a, 2010b), or only onto the radial tuberosity in P. cancrivorus (Pereira et al. 2010). In non-procyonid caniforms, the proximal origin was normal, while the inclusion of the lateral supracondylar crest as an origin was only reported in Aonyx sp. (Macalister 1873a), G. cuja (Ercoli et al. 2015), E. lutris (Howard 1973), P. pennanti (Feeney 1999), A. fulgens (Fisher et al. 2009), M. ursinus (Annie et al. 2019), and A. melanoleuca (Davis 1964). In A. melanoleuca, the origin from the lateral supracondylar crest was attributed to a second head (Davis 1964), similar to our observations in one specimen of P. flavus.

We identified discrepancies in the nomenclature used in previous studies concerning the m. triceps brachii in P. flavus. Beswick-Perrin (1871) did not describe all capita of the m. triceps brachii and reported that the m. anconeus is not segmented from the medial portion of the m. triceps brachii. In contrast, Julitz (1909) described four heads (capita) as anconeus muscles, corresponding to those we have identified. However, the muscle that we have designated as m. anconeus medialis was not described by Julitz (1909). Furthermore, the muscle could not be segmented from the caput mediale of the m. triceps brachii, as described by Beswick-Perrin (1871). Additionally, another recent study found that m. anconeus lateralis was completely independent in P. flavus (Böhmer et al. 2020), consistent with our findings. Previous studies did not describe the caput accessorium or a homologous portion in P. cancrivorus, N. nasua and N. narica (Mackintosh 1875; Windle 1888). However, a recent study (Tarquini et al. 2023) described it in the former two species. In one P. lotor (Allen 1882), the caput accessorium was described as accessory fibers from the humeral neck, and the “fourth head” actually corresponded to m. anconeus medialis (Table 3). Böhmer et al. (2020) described the m. anconeus medialis as the caput accessorium of the m. triceps brachii, and the caput accessorium was included on the caput mediale in caniforms. However, based on their figures, the caput accessorium is separated from the caput mediale as found in all procyonids (Allen 1882; Windle and Parsons 1897; Beddard 1900; Davis 1949; Feeney 1999; Tarquini et al. 2023). Therefore, in procyonids, the m. triceps brachii typically has four capita, similar to most caniforms. Additionally, other authors found an extra caput laterale originating from the lateral supracondylar crest in P. lotor (Windle and Parsons 1897), corresponding to the observations from one of our P. flavus especimens (M1). In other caniforms, additional heads have been reported. A total of five capita were reported due to the presence of another long head, such as the “posterior caput longum” in some ursids (Davis 1949, 1964; Annie et al. 2019), the “caput magnum” in the canid Lycaon pictus (Smith et al. 2020), and the “caput anguli” in some mustelids (Windle and Parsons 1897; Hall 1926; Ercoli et al. 2015). Other authors also reported in mustelids this additional caput as the m. tensor fasciae antebrachii, the m. epitrochelaris (Table 3), or as one part of it (Macalister 1873b; Leach 1977; Feeney 1999; Böhmer et al. 2018, 2020; Yousefi et al. 2018). In the mustelid T. taxus, this caput was designated as “triceps brachii-long head”, while the caput longum was termed as “triceps brachii – medial head” (Moore et al. 2013) (Table 3). Considering that these authors described the humeral capita as a single caput (“Triceps brachii – lateral head” Moore et al. 2013), which may indicate an incorrect description of the m. triceps brachii in T. taxus.

Concerning the m. tensor fasciae antebrachii, our study identified variations in its configuration among the examined caniforms. The muscle was found to be divided into two parts, with diverse origins for both the caudal and cranial parts. Previous descriptions did not describe this division and offered different configurations for its origins. Only in the case of P. lotor, the muscle was also reported as divided into two parts (Feeney 1999).. In the case of P. flavus, the caudal part was described as a slip originating from the m. cutaneus trunci (Beswick-Perrin 1871) or the m. teres major (Julitz 1909). The cranial part was reported as originating from the m. latissimus dorsi (Beswick-Perrin 1871; Julitz 1909), corresponding with our descriptions. Moreover, in P. flavus, the muscle was described as only originating from the m. latissimus dorsi (Mackintosh 1875; Allen 1882; Santos et al. 2010a, 2010b; Böhmer et al. 2020), as similarly reported for other procyonids. In N. nasua, the muscle also originated from the m. cutaneus trunci, while in P. cancrivorus, it originated from two parts of the m. latissimus dorsi (Tarquini et al. 2023). In B. alleni, the m. tensor fasciae antebrachii was only reported as a slender and strap-shaped muscle with no descriptions fir its attachments (Beddard 1900). Typically, an unique insertion onto the olecranon was described in most procyonids (Mackintosh 1875; Allen 1882; Santos et al. 2010a; Böhmer et al. 2020), while insertions onto the olecranon and antebrachial fascia were also reported in several procyonids, including P. flavus (Beswick-Perrin 1871; Julitz 1909), N. nasua (Tarquini et al. 2023), and P. cancrivorus (Pereira et al. 2010; Tarquini et al. 2023). In the case of B. alleni, an additional muscle was found originating from the tendinous junction of the latissimus dorsi and teres major muscles, and inserting nearer to the elbow joint than to the m. tensor fascia antebrachii (Beddard 1900). In P. cancrivorus, a belly was found caudal to the m. tensor fasciae antebrachii originating from m. cutaneus trunci (Tarquini et al. 2023). It was not considered a part of the m. tensor fasciae antebrachii due to the separation of both parts by fascia (Tarquini et al. 2023). However, based on the belly topology and previous dissections performed on the distribution of the radial nerve in P. cancrivorus specimens (Vélez García et al. 2023), this part was also innervated by the branch to the m. tensor fasciae antebrachii (Fig. S2). Therefore, the other bellies of P. cancrivorus and B. alleni could be considered homologous to the caudal part of the m. tensor fasciae antebrachii of P. flavus (Table 3).

In non-procyonid caniforms, the m. tensor fasciae antebrachii has also been reported as consisting of two parts in the ursid A. melanoleuca (Davis 1964), as well as in several mustelids, including E. barbara (Macalister 1873b), M. caurina (Hall 1926), M. martes (Yousefi et al. 2018), and P. pennanti (Feeney 1999). Notably, one part originates from m. cutaneus trunci in E. barbara (Macalister 1873b) and A. melanoleuca (Davis 1964). However, the muscle may have three portions in U. americanus, which originate from the m. teres major, the m. cutaneus trunci, and the caudal margin of the scapula (Shepherd 1883). In contrast, the muscles or a homologous portion were not reported in U. maritimus (Kelley 1888). Among mephitids, the muscle is represented by muscle fibers of the caput longum of the m. triceps brachii, which originate from the medial belly of the m. latissimus dorsi (Hall 1926). The caudal part of this muscle is only present in certain mustelids, such as M. foina (Böhmer et al. 2018), M. martes (Böhmer et al. 2018), M. americana (Leach 1977), P. pennanti (Leach 1977), and E. lutris (Howard 1973). This portion was considered the caput anguli of the m. triceps brachii because it originates from the caudal angle of the scapula and inserts directly onto the olecranon in some mustelids (Windle and Parsons 1897; Hall 1926; Ercoli et al. 2015) and mephitids (Hall 1926). In the mustelid T. taxus, this portion was designated as the caput longum of the m. triceps brachii, and the cranial part as the “epitrochlearis” originating from the latissimus dorsi and pectoralis muscles (Moore et al. 2013). Our own dissection of a Lontra longicaudis specimen revealed that the muscle presented a similar arrangement to that described in E. lutris (Howard 1973) and G. cuja (Ercoli et al. 2015). Additionally, the muscle was innervated by the first branch of the radial nerve (Fig. S2), similar to the innervation found in procyonids (Enciso-García and Vélez-García 2022; Vélez García et al. 2023). Consequently, it is possible that the caput angulare of the m. triceps brachii, as designated by the aforementioned authors in mustelids and mephitids (Windle and Parsons 1897; Hall 1926; Ercoli et al. 2015), may indeed be the m. tensor fasciae antebrachii, or a portion thereof (Ercoli et al. 2015). In the ailurid A. fulgens, the muscle may exhibit two parts, with the cranial part originating from the m. latissimus dorsi and the caudal part from the m. teres major, as reported by Carlsson (1925). It may also exhibit an indivisible muscle that originates from a raphe situated between the latissimus dorsi and teres major muscles, and fuses with the m. cutaneus trunci (Fisher et al. 2009). Among canids, only the cranial part appears to be present (Feeney 1999; Vélez-García et al. 2018b; Hermanson 2020; Smith et al. 2020), with the muscle being entirely absent in Cuon alpinus and Vulpes vulpes (Böhmer et al. 2020).

In other studies of P. flavus, the m. anconeus medialis was referred to as anconeus epitrochlearis (Beswick-Perrin 1871; Windle and Parsons 1897). However, our study did not corroborate certain described characteristics, such as insertion onto the anconeus process and the presence of continuous fibers connecting to the m. flexor carpi ulnaris (Beswick-Perrin 1871). In previous studies about P. flavus, this muscle was either left unnamed (Julitz 1909) or erroneously labeled as the caput accessorium mediale of the m. triceps brachii (Böhmer et al. 2020). In N. narica, the anconei muscles are united to the biceps, while the m. anconeus medialis originates from the supracondylar foramen in N. nasua (Mackintosh 1875; Tarquini et al. 2023). In B. alleni, only the muscle itself was reported (Beddard 1900). In other studies, the m. anconeus medialis or a homologous portion was not described in P. cancrivorus (Windle 1888; Pereira et al. 2010; Santos et al. 2010b) and N. nasua (Santos et al. 2010a). However, recent findings in both species (Tarquini et al. 2023; Vélez García et al. 2023) indicate similar attachments to those observed in P. flavus. Furthermore, this muscle is innervated by the ulnar nerve (Vélez García et al. 2023), corresponding to innervation patterns observed in P. flavus (Windle and Parsons 1897; Enciso-García and Vélez-García 2022) and P. lotor (Windle and Parsons 1897). This corroborates that the muscle is derived from the m. flexor carpi ulnaris rather than the m. triceps brachii (Diogo and Abdala 2010; Vélez García et al. 2023). It is important to highlight that the muscle is constantly present in most caniform families, with exceptions observed in the Canidae and Ursidae families. Nevertheless, it may be fused to the caput mediale of the m. triceps brachii in the ursid A. melanoleuca (Davis 1964), or it may be vestigial in U. maritimus (Kelley 1888) and the canid C. thous (Vélez-García et al. 2018a) (Table 4). Additionally, although in the canid C. alpinus, Böhmer et al. (2020) reported this muscle, their figures may indicate a vestigial behavior as similarly found in C. thous (Vélez-García et al. 2018a).

The shoulder movements play a pivotal role in the arboreal and prehensile habits of P. flavus, requiring high abilities and precise control to move among the tree branches. Although the extrinsic muscles contribute to shoulder support, the intrinsic muscles execute more precise shoulder joint movements. Further, the scapular muscles are responsible for the lateral and medial rotation of the humerus when the limb is free. These muscles also stabilize the shoulder joint when the limb is in contact with a substrate. These functions are not unique to P. flavus but are also observed in other caniforms (Fisher et al. 2009; Moore et al. 2013; Souza-Junior et al. 2018; Vélez-García et al. 2018b; Smith et al. 2020). The m. teres major in P. flavus plays a critical role in flexing the shoulder, working synergistically with the m. latissimus dorsi (Vélez-García and Miglino 2023). It should facilitate the medial rotation of the humerus when the limb is not resting on a substrate. These actions are essential for P. flavus to propel its body trunk cranially during vertical tree climbing or when is navigating between branches. Additionally, the presence of a m. coracobrachialis longus with a more distal insertion onto the humerus increases adduction strength in the shoulder (Kardong 2012; Monroy-Cendales et al. 2020). This capability allows P. flavus to keep vertical suspension on tree trunks while climbing. Therefore, the m. coracobrachialis longus provides greater support when medially pulling the thoracic limb with more strength, while contracting the digital flexor muscles to grip the branches. In this case, taxa with this muscle (P. flavus, some mustelids, A. fulgens, and ursids) must have advantages over other specimens when climbing trees.

The m. biceps brachii in P. flavus delivers similar functions to those of other caniforms, including elbow flexion, shoulder extension, and cranial and medial stabilization of the shoulder joint (Moore et al. 2013; Souza-Junior et al. 2018; Vélez et al. 2018; Smith et al. 2020). Furthermore, the larger origin of the m. brachialis that reaches the lateral supracondylar crest, as observed in most arctoids, enhances elbow flexion. Elbow flexion is also supported by the brachioradialis and extensor carpi radialis muscles, mainly when the limb already exhibits some degree of flexion (Vélez-García et al. 2022; Vélez-García et al. 2022). The m. biceps brachii was previously considered a supinator in functional studies of the thoracic limb muscles and bones in carnivorans without caput breve (Taylor 1982; Feeney 1999), recent studies have not described this function (Fisher et al. 2009; Ercoli et al. 2015; Taverne et al. 2018; Böhmer et al. 2019; Tarquini et al. 2023). In mammals where the m. biceps brachii has the caput breve, the muscle indeed contributes to supination (Taylor 1978; Monroy-Cendales et al. 2020, 2023; Standring 2020; Richards et al. 2023). Thus, the caput breve likely supports supination alongside the brachioradialis and supinator muscles in P. flavus and other species, such as A. fulgens and ursids of the genera Ursus and Tremarctos. Even the greater development of the m. biceps brachii may also account for the larger size of the radial tuberosity in P. flavus compared to other procyonids (Tarquini et al. 2019). In addition, P. flavus exhibits more branches from the brachial artery supplying this muscle than cursorial caniforms as C. thous (Vélez et al. 2018) and C. lupus familiaris (Hermanson et al. 2020). Therefore, this suggests increased activity of the m. biceps brachii in P. flavus.

The presence of m. coracobrachialis longus and the caput breve of m. biceps brachii in arctoid species may be phylogenetically associated with a common ancestor with proficient arboreal and prehensile abilities. Among procyonids, both muscle bellies are preserved in P. flavus, whereas the muscle probably disappeared in the common ancestor of other procyonid genera. Notably, the genus Bassaricyon, which shares similar locomotor skills with P. flavus, is the only genus showing such abilities. Therefore, although the presence of both muscle bellies is not essential for arboreal and prehensile abilities, its absence may represent an intra- and interspecific disadvantage. Moreover, the presence of both muscles likely contribute to better anatomical adaptations in most P. flavus specimens. However, the B. alleni and P. flavus specimens lacking these bellies may exhibit reduced supination and elbow flexion forces compared to other specimens, implying decreased abilities. Thus, further myological studies in B. alleni are required, as only one specimen has been studied to date (Beddard 1900), and their absence may represent an anatomical variant, similar to that observed in one P. flavus specimen (M2).

Elbow extension in caniforms is primarily performed by the caudal group of brachium muscles, including the m. triceps brachii with its four heads, m. tensor fasciae antebrachii, and m. anconeus lateralis (Souza-Junior et al. 2018; Vélez et al. 2018). Depending on the species, elbow extension may also involve an additional caput longum of the m. triceps brachii, a duplicated m. tensor fasciae antebrachii, and a m. anconeus medialis. In P. flavus, this extension is primarily achieved by the latter two muscular features. The caudal part of the m. tensor fasciae antebrachii is consistently present and well-developed in mustelids and mephitids, where it supports both shoulder and elbow flexion due to its direct origin from the scapula. In contrast, it contributes to shoulder flexion with less force in other arctoids due to its origin from soft tissues. Conversely, the cranial part of the m. tensor fasciae antebrachii only supports elbow extension and is invariably present in canids, procyonids and ursids. Therefore, the caudal part of the m. tensor fasciae antebrachii has phylogenetic significance within the infraorder Arctoidea, while the cranial part is characteristic of the suborder Caniformia.

The presence of m. anconeus medialis appears consistent in species within the infraorder Arctoidea, with the exception of ursids, where the muscle may appear in a vestigial shape or fused with the caput mediale of the m. triceps brachii (Kelley 1888; Davis 1964). This variation may be associated with the reduction (Ailuropoda and Tremarctos) or total absence (Ursus) of the supracondylar foramen in ursids (Davis 1964; Meloro and de Oliveira 2019; Jiangzuo and Flynn 2020). In canids, which lack this foramen entirely, the muscle is also completely absent or occasionally appearing in a vestigial manner (Vélez-García et al. 2018a). Consequently, in non-ursid arctoids, this muscle not only extends the elbow joint but also serves to medially stabilize it. This adaptation is particularly advantageous for smaller species that perform faster movements, requiring enhanced elbow stabilization. Meanwhile, the m. anconeus lateralis is responsible for extension and lateral stabilization of the elbow joint, with variations in force due to a more proximal origin. Consequently, in P. flavus, both anconei muscles appear to have evolved to increase elbow joint stabilization when the antebrachial muscles are working on arboreal and prehensile activities.

The presence of the caput breve of m. biceps brachii, m. coracobrachialis longus and m. anconeus medialis is not exclusive to caniforms but extends to other mammalian taxa, including monotremes (Gambaryan et al. 2015), marsupials (Diogo et al. 2016; Richards et al. 2023), anteaters (Taylor 1978; Vélez-García et al. 2020; Vélez-García et al. 2021), rodents (Böhmer et al. 2020), and non-human primates (Diogo and Wood 2012; Monroy-Cendales et al. 2020, 2023). However, the presence or absence of these muscles also varies within species of the same clade. For instance, marsupials like Vombatus (common wombat) and Didelphis (opossum), with fossorial and arboreal habits, respectively, do not have m. coracobrachialis longus (Diogo et al. 2016; Richards et al. 2023), while the arboreal marsupial Phascolarctos cinereus (koala) has it (Richards et al. 2023). In the case of semiarboreal anteaters, the genus Tamandua and the terrestrial genus Myrmecophaga feature only m. coracobrachialis longus, while the arboreal Cyclopes lacks any coracobrachialis muscle (Taylor 1978; Vélez-García et al. 2020). Similarly, saltatorial and arboreal rodents like Meriones unguiculatus and Sciurus vulgaris, respectively, preserve the caput breve of the m. biceps brachii, while the terrestrial rodent Rattus norvegicus does not possess it (Böhmer et al. 2020). Therefore, the presence or absence of these intrinsic shoulder and brachial muscles can suggest a phylogenetic significance in mammals (Diogo and Abdala 2010).

In carnivorans within the suborder Caniformia, the presence or absence of intrinsic shoulder and brachial muscles holds both phylogenetic and functional implications. Notably, canids form a distinct group independent of arctoids, indicating the divergence of these species with fewer muscle bellies to provide medial support to the shoulder and elbow. This divergence may be attributed to the fact that canids predominantly require shoulder and elbow movements in a sagittal plane (Feeney 1999; Souza-Junior et al. 2018), setting them apart from arctoids (Davis 1964). An exception is found in C. alpinus, which is located outside the canid cluster. However, if the m. anconeus medialis is considered absent due to its vestigial shape, this species would fit within the canid cluster. This vestigial presence of the muscle in canids might be traced back to a shared phylogenetic link with the common ancestor of caniforms, suggesting that medial muscular support for the elbow through the m. anconeus medialis is dispensable in canids.

Another cluster featuring fewer muscles is the fifth cluster, including mephitids, the mustelid T. taxus and mustelids of the subfamily Lutrinae. These species lack the coracobrachialis muscles and, in some cases, even the m. teres minor in mephitids and lutrines. Notably, one limb of Aonyx sp. is grouped in the fourth cluster due to the presence of m. coracobrachialis brevis. Therefore, if the tendency of lutrines had been taken into account, this limb would have been classified into the fifth cluster. This indicates that these species require less intrinsic muscular support for the shoulder. Thus, the coracobrachiales muscles may appear in a vestigial shape or fuse with another muscle. The absence of these muscles in mephitids could be a feature that emerged during their divergence from the common ancestor of ailurids and mustelids, while in lutrines, it might have occurred during their subsequent divergence within the family Mustelidae. The consistent presence of the m. teres minor in T. taxus indicates a phylogenetic connection with non-lutrine mustelids.

In the case of P. flavus, the intrinsic shoulder and brachial muscles consistently maintain a phylogenetic association within the infraorder Arctoidea, primarily with the ailurid A. fulgens and the ursids of the genera Ursus and Tremarctos (Cluster 3). This implies that these species have the highest level of muscular support for shoulder and elbow movements, including enhanced supination compared to other specimens and species. In the first and fourth clusters, support for elbow flexion and supination is diminished due to the absence of the caput breve of the m. biceps brachii. Within the first group, the enhanced medial support for the shoulder results from the presence of m. corabrachialis longus. Consequently, some P. flavus specimens have reduced intrinsic muscle support for the shoulder and elbow, similar to other procyonids, mustelids, and the ursids A. melanoleuca and M. ursinus. Nevertheless, as previously discussed, this may indicate functional disadvantages for specimens preserving all these muscles. Additionally, individual specimens may exhibit dominance of one thoracic limb, as observed in PfS4, where the right limb belongs to the first cluster, while the left limb is grouped within the third cluster.

In summary, P. flavus stands as the sole extant musteloid species potentially preserving the intrinsic shoulder and brachial muscles traced back to the common ancestor of arctoids. Furthermore, the presence of all these muscles in caniforms appears to provide both intra- and interspecific functional advantages. Consequently, P. flavus specimens equipped with all intrinsic shoulder and brachial muscles may exhibit greater strength and enhanced abilities in the shoulder and elbow when compared to other specimens or different species.

Despite our findings, it is important to acknowledge certain limitations in this study. Many authors did not provide detailed anatomical descriptions and often analyzed few specimens. Additionally, some characteristics were assigned based on family tendencies introducing potential biases into the statistical analysis. Therefore, to improve future investigations, it is imperative to conduct detailed anatomical studies including a broader range of arctoid species, using a larger sample size. Finally, although this study focused on a specific group of muscles in caniforms, future comparative analyses applying similar statistical methods can extend to diverse muscle groups and species.