Ophiacodon retroversus

The reconstruction presented in Figure 2 is based principally on FMNH UC 458, a well-preserved partial skeleton that was extensively featured by Romer and Price (1940). Other specimens in the FMNH and MCZ collections contributed additional information, especially regarding finer-scale nuances of muscle attachments. Inferred origins and insertions are described in Table 1. Ophiacodon was codified as part of Ophiacodontidae, and reconstructed states at the node Metopophora ( Table S1 (Tables_S1-8.xlsx)) were used to clarify remaining uncertainties. Some character states remained ambiguous for this node, and these were resolved for Ophiacodon as follows.

Figure 2.

Reconstructed muscle attachments in Ophiacodon. A, pelvis and axial skeleton, in lateral, posterior, and anterior views. B, femur in anterior, lateral, posterior, and medial views. C, tibia and fibula in anterior, lateral, posterior, and medial views. D, pes in dorsal (top) and ventral (bottom) views. Red denotes inferred origins, blue denotes inferred insertions, and orange denotes ambiguous attachments (see text). Scale = 10 mm. See Table 1 for muscle abbreviations.

Table 1.

Reconstructed origins and insertions in Ophiacodon retroversus. LOI, level of inference in the scheme of Witmer (1995).

Table 1.

Continued.

Table 1.

Continued.

Character 12. The iliac blade of ophiacodontids is small, with a restricted area ventral to the transverse line for origin of the IF, supporting reconstruction of a single head in Ophiacodon (state 0, level II′ inference in the scheme of Witmer, 1995).

Character 15. Without any positive evidence for a splitting of the muscle mass, the PIFI is conservatively reconstructed as singular here (state 0, level II′ inference). It would be anticipated that, from a functional perspective, two separate heads with highly similar lines of action would not be substantially different from a singular head, assuming that their combined mass and force-generating capacity were the same as that of the singular head.

Character 16. The plesiomorphic, anteroposteriorly expansive puboischiadic plate, and minimal preacetabular extension of the iliac blade, strongly supports reconstructing the PIFI origin in the primitive state in all “pelycosaurs” (state 0, level II′ inference).

Character 18. There is no positive evidence for a splitting of the muscle mass in “pelycosaurs.” and conservatively a single head would be reconstructed. However, the inferred origin (character 19) occupies an extensive area, rendering it more likely that some level of anteroposterior subdivision existed, at least in function if not actual anatomy (state 1, level II′ inference).

Character 19. Much like character 16, the plesiomorphic morphology of the puboischiadic plate suggests the PIFE existed in the primitive state in all “pelycosaurs” (state 0, level II′ inference).

Character 27. The FTI2 is synapomorphic for Amniota (character 23), and so it can be inferred to have existed in Ophiacodon. As hypothesized by Bishop and Pierce (2024b), the persistence of a fibular tubercle suggests that the switch from ILFB-dominated (TEN-dominated) to FTI2-dominated (BICF-dominated) attachment on the fibula probably occurred crownward of Mammaliaformes. An insertion on the tibia is therefore reconstructed here (state 1, level II′ inference).

Character 28. The presence of the FTE is most parsimoniously inferred as a synapomorphy of Amniota (or a more inclusive clade), and only a single head is conservatively assumed here for the FTE (state 1, level II′ inference).

Character 29. The FTE was present in “pelycosaurs” (see previous character), but no osteological correlate of its origin is present (i.e., not states 0 or 3). The locus of its attachment would have been somewhere between the posterior ilium and dorsal ischium, and it is tentatively reconstructed in the middle of this area, spanning between the two bones via soft tissues (state 1, level II′ inference). A dorsoventrally shallow posterior ilium in Ophiacodon further argues against any attachment to this bone. In functional terms, the exact dorsal position of origin would only be expected to have a potentially important effect on the capacity for hip elevation (limb abduction) and depression.

Character 30. The posterior proximal tibia of large, well-preserved Ophiacodon specimens bears a characteristic linear region of rugose scarring (Bishop and Pierce, 2024b). The scars are always directed proximally, suggesting that they denote insertion of either one or more of the long flexors (excluding the PIT), or the POP. A feature arguing against the latter is the inclination of the linear region, such that the scars are not well suited for receiving a muscle spanning from the fibula laterally. It is not possible to identify which long flexor(s) inserted here, but the FTE usually is one of the more laterally inserting of the muscles in extant species, and as the dorsalmost one in origin, it is topographically “easiest” to reconstruct as inserting here in extinct forms such as Ophiacodon. A FTE insertion is tentatively inferred here (level II inference).

Character 31. The FTE can be confidently reconstructed as present in Ophiacodon, and in light of the lack of any positive osteological evidence, the conservative assumption is that a secondary attachment to the GL was absent (state 2, level II′ inference).

Character 34. “Pelycosaurs” are inferred to have retained the plesiomorphic configuration of the PIFI mass (characters 15 and 16), which in turn implies that the PUT also existed in its primitive, single-bellied condition (state 0, level II′ inference).

Character 35. Following the same logic as for the preceding character, a singular insertion is reconstructed for the PUT (state 0, level II′ inference).

Character 37. As the ADD is unambiguously reconstructed as possessing a single head (character 36), this obviates the possibility of a crocodylian-like origin. However, it is not possible to discern between the remaining potential states in the absence of osteological evidence; whether or not it gained direct osteological attachment to the pubis and/or ischium is uncertain. A conservative, middle-of-the-road approach is taken here, and the origin is reconstructed as close to the puboischiadic suture on the ventrolateral puboischiadic plate, which is topographically intermediate between states 1 and 3 (level II′ inference). Whether a direct osteological attachment was present or not would have negligible effect in functional terms.

Character 38. Phylogenetic optimization recovers state 1 as unambiguous for Metopophora; however, Ophiacodon retroversus itself codes as state 2 here.

Character 43. The tail of “pelycosaurs” is extensive, although the region typically associated with the CF origin in extant long-tailed sauropsids (long head) is notably shorter in “pelycosaurs” (Bishop and Pierce, 2024b); for example, transverse processes are completely absent by the 11th caudal vertebra in MCZ VPRA-1366 Ophiacodon uniformis. The locus of origin in tetrapods is primitively an osseous attachment somewhere in the ventrolateral quadrant of the proximal tail, but exactly where direct attachment occurred, and whether this included the posterior ilium, is uncertain. The origin is tentatively inferred to be restricted to just the proximal caudal vertebrae in Ophiacodon (state 0, level II′ inference).

Character 46. The gastrocnemii are ubiquitous among amniotes, and so the GE can be confidently inferred as present in “pelycosaurs.” Absence of a calcaneal tuber, fabellae, or parafibula in “pelycosaurs,” coupled with the plesiomorphic “flattened” nature of the tarsals, suggests that the GE most likely existed in the primitive, single-headed condition in Ophiacodon (state 1, level II′ inference).

Character 48. As per character 46, a distinct PLA is considered absent in Ophiacodon (state 0, level II′ inference).

Character 50. The (at least partial) involvement of the posterior to medial proximal tibia in the GI origin is widespread across amniotes. The posteromedial proximal tibia area is here conservatively reconstructed as the primary locus of the origin for the GI in Ophiacodon (state 1, level II′ inference), but an attachment to the distal femur is not implausible.

Character 51. The GE can be inferred as present in Ophiacodon, but the absence of a calcaneal tuber implies the plesiomorphic, aponeurotic form of insertion (state 1, level II′ inference).

Character 52. As per character 46, a distinct PLA is considered absent in Ophiacodon (state 0, level II′ inference).

Character 54. As per character 51, a common aponeurotic insertion with the GE is reconstructed (state 1, level II′ inference).

Character 55. The locus of the FDLh origin in amniotes is generally centered around the posterior proximal fibula, and this area is conservatively inferred to be the primary attachment for the FDLh in Ophiacodon (state ∼3, level II′ inference; parafibula is unknown in “pelycosaurs”). However, it cannot be ruled out that the origin extended more proximally to include the distal femur.

Character 62. In the absence of any positive osteological evidence, the plesiomorphic state is reconstructed for Ophiacodon here (state 0, level II′ inference).

Character 64. Plesiomorphically, the general locus of origin for the EDBh is the ankle joint, but extant mammals exhibit an origin near the proximal fibula; the origin presumably underwent proximal migration on the line to mammals. The EDBh in Ophiacodon is conservatively inferred to originate from the anterior aspect of the distal fibula (state 2, level II′ inference), spatially midway between the mammalian and nonmammalian conditions.

Character 66. The EHL is ubiquitous among amniotes and can be confidently inferred as present in “pelycosaurs.” Its locus of origin is centered around the anterior fibula, and in the absence of any positive evidence, a middle-of-the-road approach is taken, reconstructing the muscle's origin in Ophiacodon at the fibular midshaft (state 2, level II′ inference).

Character 69. An insertion of the TA distal to the ankle is ubiquitous among amniotes, with the general locus of attachment located at metatarsal I. A more plesiomorphic insertion on the anteromedial proximal aspect of metatarsal I is reconstructed here (state 1, level II′ inference), which is consonant with the more “reptilian” construction of the pes in “pelycosaurs,” namely, a primitive tarsal configuration, broad and flat proximal tarsals, and strong asymmetry in digit lengths.

Character 72. Two peroneal heads can be confidently reconstructed (character 70), permitting reconstruction of the PB as originating from the anterolateral proximal fibula in Ophiacodon (state 3, level II′ inference).

Character 74. Two peroneal heads can be confidently reconstructed (character 70), with the PB inserting in a mammal-like fashion somewhere along pedal ray V. An insertion on metatarsal V is inferred here (state 4, level II′ inference).

Character 75. Strictly speaking, unambiguous osteological evidence is wanting for this character in ophiacodontids, and phylogenetic inference results in state 0 being reconstructed here. However, the lateral tibial shaft possesses a distinctive line of fibrous or rugose scarring, with the scars “trending” distally, and these scars in some specimens are clearly separate from the rugose scarring proximally inferred as the attachment of the POP. A similar condition is observed in some other “pelycosaurs” as well (Bishop and Pierce, 2024b, fig. 29). This may denote a tibial attachment for the PP (state 1, level II inference), or alternatively for the IC.

Character 76. The plesiomorphic condition of the “pelycosaur” pes suggests that a more mammal-like insertion of the PP is unlikely; in any case, the locus of insertion was probably centered on the plantar aspect of the medial pes. A primary attachment to the medial distal tarsals is inferred here (state 2, level II′ inference).

Dimetrodon milleri

The reconstruction presented in Figure 3 is based principally on the holotype, MCZ VPRA-1365, a near-complete individual (Romer and Price, 1940), although other specimens of various Dimetrodon species in the MCZ, FMNH, and AMNH collections furnished additional information. Inferred origins and insertions are described in Table 2. Dimetrodon was codified as part of a paraphyletic “Sphenacodontia” (see Bishop and Pierce, 2024b), and reconstructed states at the node Sphenacodontia sensu stricto ( Table S2 (Tables_S1-8.xlsx)) were used to clarify remaining uncertainties. Some character states remained ambiguous for this node, and these were resolved for Dimetrodon as follows.

Figure 3.

Reconstructed muscle attachments in Dimetrodon. A, pelvis and axial skeleton, in lateral, posterior, and anterior views. B, femur in anterior, lateral, posterior, and medial views. C, tibia and fibula in anterior, lateral, posterior, and medial views. D, pes in dorsal (top) and ventral (bottom) views. Red denotes inferred origins, blue denotes inferred insertions, and orange denotes ambiguous attachments (see text). Scale = 10 mm. See Table 2 for muscle abbreviations.

Table 2.

Reconstructed origins and insertions in Dimetrodon milleri. LOI, level of inference in the scheme of Witmer (1995).

Table 2.

Continued.

Table 2.

Continued.

Character 12. The iliac blade of sphenacodontians remains of modest size, conservatively suggesting that only a single head was still present in Dimetrodon (state 0, level II′ inference). However, it is notably expanded relative to more basal “pelycosaurs,” suggesting that the IF at least had a larger origin, if not physically subdivided (character 13; see Bishop and Pierce, 2024b).

Character 15. Without any positive evidence for a splitting of the muscle mass, the PIFI is conservatively reconstructed as singular here (state 0, level II′ inference).

Character 16. The plesiomorphic, anteroposteriorly expansive puboischiadic plate, and minimal preacetabular extension of the iliac blade, strongly supports reconstructing the PIFI origin in the primitive state in all “pelycosaurs” (state 0, level II′ inference).

Character 18. There is no positive evidence for a splitting of the PIFE mass in “pelycosaurs,” and conservatively a single head would be reconstructed. However, as explained above for Ophiacodon, it is probable that some level of anteroposterior subdivision existed (state 1, level II′ inference).

Character 19. Much like character 16, the plesiomorphic morphology of the puboischiadic plate suggests the PIFE existed in the primitive state in all “pelycosaurs” (state 0, level II′ inference).

Character 27. Following the same logic as outlined above for Ophiacodon, an insertion on the tibia is reconstructed for Dimetrodon (state 1, level II′ inference).

Character 28. The presence of the FTE is most parsimoniously inferred as a synapomorphy of Amniota (or a more inclusive clade), and only a single head is conservatively inferred for the FTE (state 1, level II′ inference).

Character 29. The FTE was present in “pelycosaurs” (see previous character), but no osteological correlate of its origin is present (i.e., not states 0 or 3). The locus of its attachment would have been somewhere between the posterior ilium and dorsal ischium, and it is tentatively reconstructed in the middle of this area, spanning between the two bones via soft tissues (state 1, level II′ inference).

Character 30. The posterior proximal tibia of large, well-preserved Dimetrodon specimens (especially of limbatus and gran-dis species) bears a broad, diffuse region of notably more rugose bone texture, sometimes quite similar to that which is characteristic of large Ophiacodon specimens (Bishop and Pierce, 2024b). Following the same logic for Ophiacodon, a FTE insertion is inferred here (level II inference).

Character 31. The FTE can be confidently reconstructed as present in Dimetrodon, and in light of the lack of any positive osteological evidence, the conservative assumption is that a secondary attachment to the GL was absent (state 2, level II′ inference).

Character 34. “Pelycosaurs” are inferred to have retained the plesiomorphic configuration of the PIFI mass (characters 15 and 16), which in turn implies that the PUT also existed in its primitive, single-bellied condition (state 0, level II′ inference).

Character 35. Following the same logic as for the preceding character, a singular insertion is reconstructed for the PUT in Dimetrodon (state 0, level II′ inference).

Character 37. Following the same argument as outlined above for Ophiacodon, the ADD origin is inferred as close to the puboischiadic suture on the ventrolateral puboischiadic plate, which is intermediate between states 1 and 3 (level II′ inference).

Character 43. The tail of “pelycosaurs” is extensive, although the region typically associated with the CF origin in extant long-tailed sauropsids is notably shorter in “pelycosaurs”; for example, transverse processes are completely absent by the 11th caudal vertebra in MCZ VPRA-1365. The locus of origin in tetrapods is primitively an osseous attachment somewhere in the ventrolateral quadrant of the proximal tail, but exactly where direct attachment occurred, and whether this included the posterior ilium, is uncertain. The origin is tentatively inferred to have been restricted to just the proximal caudal vertebrae in Dimetrodon (state 0, level II′ inference).

Character 46. The gastrocnemii are ubiquitous among amniotes, and so the GE can be confidently inferred as present in “pelycosaurs.” Absence of a calcaneal tuber, fabellae, or parafibula in “pelycosaurs,” coupled with the plesiomorphic plate-like geometry of the tarsals, suggests that the GE most likely existed in the primitive, single-headed condition in Dimetrodon (state 1, level II′ inference).

Character 48. As per character 46, a distinct PLA was not present in Dimetrodon (state 0, level II′ inference).

Character 51. The GE can be inferred as present in Dimetrodon, but the absence of a calcaneal tuber implies the plesiomorphic, aponeurotic form of insertion (state 1, level II′ inference).

Character 52. As per character 46, a distinct PLA was not present in Dimetrodon (state 0, level II′ inference).

Character 54. As per character 51, a common aponeurotic insertion with the GE is inferred here (state 1, level II′ inference).

Character 55. The locus of the FDLh origin in amniotes is generally centered around the posterior proximal fibula, and this area is conservatively reconstructed as the primary attachment for the FDLh in Dimetrodon (state ∼3, level II′ inference; parafibula is unknown in “pelycosaurs”). As in Ophiacodon, it cannot be ruled out that the origin extended more proximally to include the distal femur.

Character 62. In the absence of any positive osteological evidence, the plesiomorphic state is assumed here for Dimetrodon (state 0, level II′ inference).

Character 64. Following the same argument as outlined above for Ophiacodon, the EDBh origin is inferred as centered on the anterior aspect of the distal fibula (state 2, level II′ inference).

Character 66. The EHL is ubiquitous among amniotes and can be confidently inferred as present in “pelycosaurs.” Its locus of origin is centered around the anterior fibula, and in the absence of any positive evidence, a middle-of-the-road approach is taken to reconstruct its origin in Dimetrodon at the fibular midshaft (state 2, level II′ inference).

Character 69. An insertion of the TA distal to the ankle is ubiquitous among amniotes, with the general locus of attachment located at metatarsal I. A more plesiomorphic insertion on the anteromedial proximal aspect of metatarsal I is inferred here (state 1, level II′ inference).

Character 72. Following the same argument as outlined above for Ophiacodon, the PB likely originated from the anterolateral proximal fibula in Dimetrodon (state 3, level II′ inference).

Character 74. Following the same argument as outlined above for Ophiacodon, the insertion of the PB is inferred to be on metatarsal V (state 4, level II′ inference).

Character 75. Following the same argument as outlined above for Ophiacodon, a tibial attachment for the PP is inferred here for Dimetrodon (state 1, level II inference).

Character 76. Following the same argument as outlined above for Ophiacodon, a primary attachment to the medial distal tarsals is inferred here (state 2, level II′ inference).

Oudenodon bainii

The reconstruction presented in Figure 4 is based principally on NHMUK PV R.4067, a very well preserved and mostly complete skeleton (Watson, 1917, 1960). Material in the BP, GPIT, and SAM collections provided additional details for parts of the pelvis that were missing. The pes of NHMUK PV R.4067 is not preserved, and as such, the pedal anatomy was based on other material, especially that of NMQR 479 Dicynodontoides recurvidens, an excellently preserved and complete specimen missing only the tip of the ungual from digit III (King, 1985). Inferred origins and insertions are described in Table 3. Oudenodon was codified as part of Anomodontia, and reconstructed states at the node Neotherapsida ( Table S3 (Tables_S1-8.xlsx)) were used to clarify remaining uncertainties. Some character states remained ambiguous for this node, and these were resolved for Oudenodon as follows.

Figure 4.

Reconstructed muscle attachments in Oudenodon. A, pelvis and axial skeleton, in lateral, posterior, and anterior views. B, femur in anterior, lateral, posterior, and medial views. C, tibia and fibula in anterior, lateral, posterior, and medial views. D, pes in dorsal (top) and ventral (bottom) views; based on Dicynodontoides. Red denotes inferred origins, blue denotes inferred insertions, and orange denotes ambiguous attachments (see text). Scale = 10 mm. See Table 3 for muscle abbreviations.

Table 3.

Reconstructed origins and insertions in Oudenodon bainii. LOI, level of inference in the scheme of Witmer (1995).

Table 3.

Continued.

Character 12. The apomorphic anterior expansion of the iliac blade in therapsids suggests enlargement and probable differentiation of the primitive iliofemoralis mass, and so two heads are conservatively reconstructed here (state 1, level II′ inference).

Character 13. This is reconstructed unambiguously as state 1 for Neotherapsida, but within dicynodonts (including Oudenodon), the preacetabular blade further expands anteriorly, coding as state 2.

Character 15. Without any positive evidence for a splitting of the muscle mass, the PIFI mass is conservatively reconstructed as singular here (state 0, level II′ inference).

Character 16. The reduction of the pubis in dicynodonts such as Oudenodon suggests a more dorsal location for the origin of the PIFI on the pubis (Bishop and Pierce, 2024b), but likely not extending onto the ilium (state 2, level II inference).

Character 18. There is no positive evidence for a splitting of the PIFE muscle mass in dicynodonts, and conservatively a single head would be reconstructed. However, in a similar fashion to the “pelycosaurs,” as the inferred origin (character 19) covers a broad area, this suggests some degree of anteroposterior subdivision, presaging the mammalian OBEX (anterior) and QF (posterior).

Character 19. Character optimization is ambiguous for Neotherapsida, but the greatly reduced and posteriorly swept pubis and enlarged obturator foramen of most dicynodonts (including Oudenodon) support state 2 (level II′ inference).

Character 27. The FTI2 is synapomorphic for Amniota (character 23), and so can be inferred to have existed in Oudenodon. As discussed above for the “pelycosaurs,” an insertion on the tibia can be reconstructed (state 1, level II′ inference).

Character 28. The presence of the FTE is most parsimoniously inferred as a synapomorphy of Amniota (or a more inclusive clade), and only a single head is conservatively assumed for the FTE here (state 1, level II′ inference).

Character 29. The FTE was present in dicynodonts (see previous character), but no osteological correlate of its origin is present (i.e., not states 0 or 3). The locus of its attachment would have been somewhere between the posterior ilium and dorsal ischium, although the reduction of the postacetabular ilium in dicynodonts suggests that the FTE probably did not attach there. The muscle's origin is tentatively reconstructed in the middle, spanning between the two bones via soft tissues (state 1, level II′ inference). In functional terms, the exact dorsal position of origin would only be expected to have a potentially important effect on hip elevation–depression capacity. Derived kannemeyeriiform dicynodonts usually possess an ischiadic tuberosity much like that of cynodonts, which may signal in those animals a reconfiguration of the crural long flexor origins (Bishop and Pierce, 2024b).

Character 31. The FTE can be confidently reconstructed as present in Oudenodon. Following the same logic as outlined for the “pelycosaurs,” a secondary attachment to the GL was probably absent (state 2, level II′ inference).

Character 34. Although dicynodonts are inferred to have a dorsally shifted PIFI origin (character 16 above), this is not homologous to the similar shift inferred to have taken place within cynodonts (Bishop and Pierce, 2024b; see also below), and so it cannot be a priori expected to have been accompanied by changes to the PUT. In the absence of evidence to the contrary, the plesiomorphic condition is conservatively reconstructed, with a single origin on or near the pubic tubercle (state 0, level II′ inference).

Character 35. Following the same logic as for the preceding character, a plesiomorphic, singular insertion is reconstructed in Oudenodon (state 0, level II′ inference).

Character 37. As the ADD is unambiguously reconstructed as possessing a single head (character 36), this obviates the possibility of a crocodylian-like origin. However, it is not possible to discern between the remaining potential states in the absence of osteological evidence. It cannot be determined whether attachment to the pubis and/or ischium was direct or indirect (via a ligament or aponeurosis). A similar approach to the “pelycosaurs” is taken here, with an origin inferred as close to the puboischiadic junction on the ventrolateral puboischiadic plate, which is intermediate between states 1 and 3 (level II′ inference).

Character 43. The greatly reduced tail of dicynodonts suggests a more mammalian type of origin for the CF. Oudenodon is tentatively reconstructed with the monotremes' condition, in which a small osseous contact remains (state 2, level II′ inference).

Character 46. The gastrocnemii are ubiquitous among amniotes, and so the GE can be confidently inferred as present in dicynodonts. Even in well-preserved, articulated skeletons of dicynodonts, there is no evidence for ossified fabellae, or a parafibula, which might signal subdivision (although Dicynodontoides possesses an incipient calcaneal tuber; King, 1985). Without any positive evidence for a splitting of the muscle mass, the GE is conservatively reconstructed as singular here (state 1, level II′ inference).

Characters 48 and 52. As per character 46, a distinct PLA was not present in Oudenodon (state 0, level II′ inference).

Character 51. Due to the presence of a calcaneal tuber in the biarmosuchian Hipposaurus, maximum parsimony reconstructs the presence of such a feature as plesiomorphic for Neotherapsida. This is obviously highly improbable, given the suite of non-eutheriodont taxa that lack any such process. Only Dicynodontoides possesses a weak, incipient tuber, hinting at some change to the nature of insertion. A middle-of-the-road state is inferred for Oudenodon, between states 1 and 3, with the GE forming an abbreviated plantar aponeurosis that only attaches to the ventral aspect of the calcaneum and metatarsals.

Character 55. The locus of the FDLh origin in amniotes is generally centered around the posterior proximal fibula; this area is conservatively reconstructed as the primary attachment for the FDLh in Oudenodon (state ∼3, level II′ inference; parafibula is unknown in non-cynodont therapsids).

Character 61. Osteological evidence is wanting for this character, but the locus of origin of the EDLh can confidently be inferred to have been somewhere around the anterolateral aspect of the knee. Two out of the three available states involve the lateral femoral condyle, and although the distal femur lacks the lateral shelf of “pelycosaurs,” the proximal fibula remains robustly developed in therapsids, lacking a spatulate profile or parafibular development (i.e., a more plesiomorphic configuration). The attachment in Oudenodon is tentatively assumed to be centered on the distal femur (approximating state 1, level II′ inference).

Character 62. In the absence of any positive osteological evidence, the plesiomorphic state is reconstructed for Oudenodon here (state 0, level II′ inference).

Character 64. Following the same logic as outlined for Ophiacodon, the EDBh in Oudenodon is inferred to originate from the anterior aspect of the distal fibula (state 2, level II′ inference). Its proximal path is inferred as along the anterior face of the fibula, rather than between fibula and tibia, to reflect the muscle approaching a more “mammalian” position.

Character 66. The EHL is ubiquitous among amniotes and can be confidently inferred as present in dicynodonts. Its locus of origin is centered around the anterior fibula, and in the absence of any positive evidence, a middle-of-the-road approach is taken, with the muscle's origin in Oudenodon reconstructed at the fibular midshaft (state 2, level II′ inference).

Character 69. An insertion of the TA distal to the ankle is ubiquitous among amniotes, with the general locus of attachment located at metatarsal I. Although the tarsal arrangement is generally difficult to assess in most anomodonts due to poor preservation or lower levels of ossification, at least some anomodonts have the same arrangement as observed in extant mammals (Brinkman, 1981; King, 1985). This is tentatively taken to indicate that Oudenodon possessed a more mammal-like attachment of the TA, on the entocuneiform in addition to metatarsal I (state 3, level II′ inference).

Character 71. The PL can unambiguously be inferred to have originated from around the lateral knee region, but the specific location cannot be discerned in therapsids. Two of the three available states involve the lateral proximal fibula, and while the fibula retains a (plesiomorphically) robust proximal construction, an origin from this region is tentatively inferred (state 3, level II′ inference).

Character 72. Two peroneal heads can be confidently reconstructed (character 70), permitting reconstruction of the PB as originating from the anterolateral proximal fibula in Oudenodon (state 3, level II′ inference).

Character 74. Two peroneal heads can be confidently reconstructed (character 70), with the PB inserting in a mammal-like fashion somewhere along pedal ray V. An insertion on metatarsal V is inferred here (state 4, level II′ inference).

Character 76. Given that at least some anomodonts possess a mammalian tarsal arrangement, this suggests that Oudenodon may have possessed a more mammal-like attachment of the PP on the plantar aspect of the proximal tarsals (state 3, level II′ inference).

Lycaenops ornatus

The reconstruction presented in Figure 5 is based principally on AMNH FARB 2240, the holotype and a generally well-preserved, near-complete skeleton (Colbert, 1948). All elements are contained within the mount on public display, which could not be disassembled for detailed study. Thus, in some cases, the inferences for Lycaenops were more heavily dependent on observations of other gorgonopsian material in BP, SAM, GPIT, UMZC, TM, BSPG, CGS, USNM, and NMQR collections, assuming gross similarity (without evidence to the contrary). Inferred origins and insertions are described in Table 4. Lycaenops was codified as part of Gorgonopsia, and reconstructed states at the node Theriodontia ( Table S4 (Tables_S1-8.xlsx)) were used to clarify remaining uncertainties. Some character states remained ambiguous for this node, and these were resolved for Lycaenops as follows.

Figure 5.

Reconstructed muscle attachments in Lycaenops. A, pelvis and axial skeleton, in lateral, posterior, and anterior views. B, femur in anterior, lateral, posterior, and medial views. C, tibia and fibula in anterior, lateral, posterior, and medial views. D, pes in dorsal (top) and ventral (bottom) views. Red denotes inferred origins, blue denotes inferred insertions, and orange denotes ambiguous attachments (see text). Scale = 10 mm. See Table 4 for muscle abbreviations.

Table 4.

Reconstructed origins and insertions in Lycaenops ornatus. LOI, level of inference in the scheme of Witmer (1995).

Table 4.

Continued.

Character 12. The apomorphic anterior expansion of the therapsid iliac blade suggests enlargement and probable differentiation of the primitive iliofemoralis mass, and so two heads are conservatively reconstructed here (state 1, level II′ inference).

Character 15. Without any positive evidence for a splitting of the muscle mass, the PIFI is conservatively reconstructed as singular here (state 0, level II′ inference).

Character 16. Although gorgonopsians, including Lycaenops, typically have an abbreviated pubis ventrally with a constricted ischiadic peduncle, it is nonetheless still anteroposteriorly extensive. The plesiomorphic state is hence reconstructed here (state 0, level II′ inference).

Character 18. There is no positive evidence for a splitting of the muscle mass in gorgonopsians, and conservatively a single head would be reconstructed. However, in a similar fashion to the “pelycosaurs,” as the inferred origin (character 19) covers a broad area, this suggests some degree of anteroposterior subdivision, presaging the mammalian OBEX (anterior) and QF (posterior).

Character 19. Theriodontia is ambiguous for this character (0, 1, 2), but gorgonopsians including Lycaenops possess an abbreviated pubis and are hence are able to be coded as state 1 (level II′ inference; see Bishop and Pierce, 2024b).

Character 20. In Lycaenops specifically, the internal trochanter is greatly reduced, although a broad patch of scarring in the same place as the trochanter (present in other gorgonopsians) is present, indicating the insertion of the PIFE (state 3, level II inference).

Character 27. The FTI2 is synapomorphic for Amniota (character 23), and so it can be inferred to have existed in Lycaenops. As outlined for Oudenodon above, an insertion on the tibia is also reconstructed here (state 1, level II′ inference).

Character 28. The presence of the FTE is most parsimoniously inferred as a synapomorphy of Amniota (or a more inclusive clade), and only a single head is conservatively assumed here (state 1, level II′ inference).

Character 29. The FTE was present in gorgonopsians (see previous character), but no osteological correlate of its origin is present (i.e., not states 0 or 3). The locus of its attachment would have been somewhere between the posterior ilium and dorsal ischium. In a similar fashion to Oudenodon, the origin is inferred to be located in the middle, spanning between the two bones via soft tissues (state 1, level II′ inference).

Character 30. The posterior proximal tibia of some well-preserved gorgonopsian specimens bears a discrete region of rugose scarring (Bishop and Pierce, 2024b). This is located in a comparable location to scarring often observed in “pelycosaurs” (including Ophiacodon and Dimetrodon), as well as some other therapsid groups, and it probably denotes the insertion of a long flexor muscle. It is not possible to identify which long flexor(s) inserted here, but following the same logic as outlined for the other taxa above, a FTE insertion is inferred here in Lycaenops (level II inference).

Character 31. The FTE can be confidently reconstructed as present in Lycaenops. Following the same logic as outlined above for other taxa, a secondary attachment to the GL was probably absent (state 2, level II′ inference).

Character 34. The PIFI in gorgonopsians is inferred to have manifested in the plesiomorphic state (character 16), and so there is no evidence to suggest that modifications to the PUT had occurred by this point in synapsid evolution. The plesiomorphic condition is reconstructed for Lycaenops, having a single origin on or near the pubic tubercle (state 0, level II′ inference).

Character 35. Following the same logic as for the preceding character, a plesiomorphic, singular insertion is inferred in Lycaenops (state 0, level II′ inference).

Character 37. Following the same logic as outlined above for Oudenodon, the ADD origin is inferred as near the anterior margin of the ventrolateral ischium, which is intermediate between states 1 and 3 (level II′ inference).

Character 38. Gorgonopsians can variably exhibit states 0 (level II′ inference) and 1 (level II inference); Lycaenops in particular has a smooth posterior femoral shaft (state 0).

Character 43. The tail of gorgonopsians is not well known, and a complete series of caudal vertebrae have yet to be described (although the tail of the basal Viatkogorgon appears to be almost complete, comprising at least 20 vertebrae; Tatarinov, 2004). Available specimens indicate that the tail, whilst relatively short in relation to the body, was still solidly constructed, with well-developed transverse processes continuing up to at least caudal vertebra 8 (Colbert, 1948; von Huene, 1950; Tatarinov, 2004; personal observation of USNM material). This suggests that the CF probably retained a more plesiomorphic origin from the caudal vertebrae only, not yet having shifted its locus anteriorly toward the sacrals (state 0, level II′ inference).

Character 46. The gastrocnemii are ubiquitous among amniotes, and so the GE can be confidently inferred as present in gorgonopsians. Even in well-preserved, articulated skeletons of gorgonopsians, there is no evidence for ossified fabellae, or a parafibula, which might signal subdivision (although several specimens possess an incipient calcaneal heel). Without any positive evidence for a splitting of the muscle mass, the GE is conservatively reconstructed as singular here (state 1, level II′ inference).

Characters 48 and 52. As per character 46, a distinct PLA was not present in Lycaenops (state 0, level II′ inference).

Character 51. As noted above for Oudenodon, maximum parsimony reconstructs the presence of a calcaneal tuber as plesiomorphic for Theriodontia, despite most noneutheriodont taxa lacking any such process. Some gorgonopsian specimens possess a weak, incipient tuber or heel, hinting at some change to the nature of insertion (Bishop and Pierce, 2024b). A middle-of-the-road state is inferred for Lycaenops, between states 1 and 3, with the GE forming an abbreviated plantar aponeurosis that only attaches to the ventral aspect of the calcaneum and metatarsals.

Character 55. The locus of the FDLh origin in amniotes is generally centered around the posterior proximal fibula; this area is conservatively reconstructed as the primary attachment for the FDLh in Lycaenops (state ∼3, level II′ inference; parafibula is unknown in non-cynodont therapsids).

Character 61. Following the same logic as outlined above for Oudenodon, the attachment in Lycaenops is assumed to be centered on the femur (approximating state 1, level II′ inference).

Character 62. In the absence of any positive osteological evidence, the plesiomorphic state is reconstructed here for Lycaenops (state 0, level II′ inference).

Character 64. Following the same logic as outlined above for Oudenodon, the EDBh in Lycaenops is reconstructed as originating from the anterior aspect of the distal fibula (state 2, level II′ inference).

Character 66. Following the same logic as outlined above for Oudenodon, the origin of the EHL in Lycaenops is reconstructed at the fibular midshaft (state 2, level II′ inference).

Character 69. An insertion of the TA distal to the ankle is ubiquitous among amniotes, with the general locus of attachment located at metatarsal I. Gorgonopsians already had achieved the tarsal arrangement observed in extant mammals (Boonstra, 1934; Broili and Schröder, 1935; Colbert, 1948; Sidor, 2022; Bendel et al., 2023), suggesting that Lycaenops likely possessed a more mammal-like attachment of the TA, on the entocuneiform in addition to metatarsal I (state 3, level II′ inference).

Character 71. Following the same logic as outlined above for Oudenodon, the PL in Lycaenops is inferred to have originated from around the proximal fibula (state 3, level II′ inference).

Character 72. Following the same logic as outlined above for Oudenodon, the PB likely originated from the anterolateral proximal fibula in Lycaenops (state 3, level II′ inference).

Character 74. Following the same logic as outlined above for Oudenodon, the PB of Lycaenops likely inserted on metatarsal V (state 4, level II′ inference).

Character 76. Given that gorgonopsians possess a mammalian tarsal arrangement, this suggests that Lycaenops may have possessed a more mammal-like attachment of the PP on the plantar aspect of the proximal tarsals (state 3, level II′ inference).

Regisaurus jacobi

The reconstruction presented in Figure 6 is based principally on the holotype, BP/ 1/5394, which insofar as the hindlimb is concerned is substantially complete and well preserved, lacking the pubis, ischium, and distal pes (Kemp, 1978). The anatomy of the pubis and ischium in the reconstruction is based on that in other well-preserved eutherocephalian specimens, including UMZC T.837 Scaloposaurus constrictus (Kemp, 1986), SAM-PK-K8798 Choerosaurus dejageri (Haughton, 1929), and BP/1/3973 Olivierosuchus parringtoni (Fourie and Rubidge, 2007). The anatomy of the distal pes is likewise informed from specimens including UMZC T.837, BP/1/4021 Scaloposaurus sp., SAM-PK-K10704 Ictidosuchoides cf. longiceps, and AMNH FARB 5622 Bauria cynops (Schaeffer, 1941). Inferred origins and insertions are described in Table 5. Regisaurus was codified as part of Eutherocephalia, and reconstructed states at the node Therocephalia ( Table S5 (Tables_S1-8.xlsx)) were used to clarify remaining uncertainties. Use of an alternative tree topology, wherein Therocephalia is paraphyletic with respect to Cynodontia (Bishop and Pierce, 2024a, 2024b; see also Pusch et al., 2024), did not change the results. Some character states remained ambiguous for this node, and these were resolved for Regisaurus as follows.

Figure 6.

Reconstructed muscle attachments in Regisaurus. A, pelvis and axial skeleton, in lateral, posterior, and anterior views. B, femur in anterior, lateral, posterior, and medial views. C, tibia and fibula in anterior, lateral, posterior, and medial views. D, pes in dorsal (top) and ventral (bottom) views; phalangeal proportions based on SAM-PK-K10704 Ictidosuchoides cf. longiceps. Red denotes inferred origins, blue denotes inferred insertions, and orange denotes ambiguous attachments (see text). Scale = 10 mm. See Table 5 for muscle abbreviations.

Table 5.

Reconstructed origins and insertions in Regisaurus jacobi. LOI, level of inference in the scheme of Witmer (1995).

Table 5.

Continued.

Character 2. Although the state is unambiguous for Regisaurus, the exact location on the ilium is ambiguous, given the uncertainty in precisely when the migration of the ITa to RF took place (Bishop and Pierce, 2024b). Therocephalian ilia are characterized by two anteriorly directed processes on the anterior-to-dorsal margin, the lower one of which is especially pronounced in Regisaurus. It remains uncertain if the development of these processes was related to appendicular or alternatively axial or costal musculature. The origin of the ITa is here reconstructed as on the anterodorsal iliac margin on the “main” part of the blade, but it is possible that it may have extended onto one of the anterior processes. In contrast, Kemp (1978) interpreted the origin in Regisaurus as near the anterior base of the ilium, rather like the mammalian RF, citing the presence of a shallow depression. However, such a structure may be a by-product of the strong development of the lower anterior iliac process in this taxon. Convincing osteological evidence for an RF-like origin in synapsids is not observed until advanced cynodonts (Bishop and Pierce, 2024b).

Character 12. As with Oudenodon and Lycaenops, the anteriorly expanded iliac surface area supports the reconstruction of two heads (state 1, level II′ inference).

Character 15. Without any positive evidence for a splitting of the muscle mass, the PIFI is conservatively reconstructed as singular here (state 0, level II′ inference).

Character 16. Eutherocephalians possess an anteroposteriorly abbreviated pubis, supporting the reconstruction of a restricted origin on the medial surface of the proximal pubis or ventral ilium (state 1, level II′ inference). Kemp (1978) reconstructed the origin on the ventral lateral aspect of the preacetabular iliac blade, in a comparable fashion to the iliac fossa of mammals; however, these structures are not phylogenetically continuous (i.e., nonhomologous), and such an inference is less well supported (state ∼4, level III inference). As noted by Bishop and Pierce (2024b), the “fossa” recognized by Kemp may instead be a spandrel of the anterior iliac processes.

Character 18. Following the same line of reasoning as for Oudenodon and Lycaenops, the PIFE likely possessed some form of anteroposterior subdivision.

Character 27. Following the same line of reasoning as for Oudenodon and Lycaenops, an insertion on the tibia is reconstructed for Regisaurus (state 1, level II′ inference).

Character 28. The presence of the FTE is most parsimoniously inferred as a synapomorphy of Amniota (or a more inclusive clade), so only a single head is inferred for the FTE here (state 1, level II′ inference).

Character 29. The FTE was present in therocephalians (see previous character), but no osteological correlate of its origin is present (i.e., not states 0 or 3). The relatively “reptilian” arrangement of the pelvis (long postacetabular iliac blade, ischium well ventral to level of acetabulum) and lack of an ischiadic tuberosity imply a more “reptilian” (state 1 or 2) origin; in the absence of any scarring on the posterior ilium or ischium, the origin is reconstructed as from the ilioischiadic ligament and fascia covering the proximal tail (state 1, level II′ inference).

Character 31. Following the same line of reasoning as for Oudenodon and Lycaenops, a secondary attachment to the GL was absent in Regisaurus (state 2, level II′ inference).

Characters 34 and 35. Eutherocephalians are inferred to have exhibited a slight shift in PIFI origin (character 16, state 1). If Therocephalia is monophyletic, such a shift would probably have been convergent with that inferred in cynodonts, since scylacosaurids retained a well-developed (i.e., anteroposteriorly long) pubic plate (Boonstra, 1964), implying the retention of a plesiomorphic PUT. However, in a scenario of Therocephalian paraphyly, where eutherocephalians are more closely related to cynodonts (see Bishop and Pierce, 2024b), a PIFI origin shift is phylogenetically congruent with the inferred shift on the stem lineage, wherein PUT reorganization in eutherocephalians becomes more likely. In the absence of other osteological evidence, it is currently not possible to discriminate between the plesiomorphic (state 0) and derived (state 2) conditions.

Character 37. As the ADD is unambiguously reconstructed as possessing a single head (character 36), this obviates the possibility of a crocodylian-like origin. However, it is not possible to discern between the remaining potential states in the absence of osteological evidence. As the puboischiadic plate is markedly transformed in eutherocephalians (paralleling the same transformation in cynodonts), the more mammalian condition is tentatively inferred in Regisaurus (state 3, level II′ inference), reconstructing the origin at the ventral junction of the pubis and ischium.

Character 43. The greatly reduced tail of eutherocephalians (Kemp, 1986; Fourie and Rubidge, 2007) suggests a more mammalian type of origin for the CF. The condition observed in monotremes is reconstructed here for Regisaurus, whereby an osseous contact is still retained (state 2, level II′ inference).

Character 46. The gastrocnemii are ubiquitous among amniotes, and so the GE can be confidently inferred as present in eutherocephalians. The development of a distinct calcaneal tuber in many baurioid eutherocephalians (including Regisaurus) may be argued to signal increasing differentiation into multiple heads, but extant crocodylians also possess a calcaneal tuber and an undivided GE. Even in well-preserved, articulated skeletons (e.g., BP/1/3973 Olivierosuchus parringtoni, SAM-PK-K10704 Ictidosuchoides cf. longiceps), there is no evidence for ossified fabellae, or a parafibula, which might signal subdivisions. Without any positive evidence for a splitting of the muscle mass, the GE is conservatively reconstructed as singular here (state 1, level II′ inference).

Character 48. As per character 46, a distinct PLA was not present in Regisaurus (state 0, level II′ inference).

Character 52. As per character 46, a distinct PLA was not present in Regisaurus (state 0, level II′ inference).

Character 55. The locus of the FDLh origin in amniotes is generally centered around the posterior proximal fibula; this area is conservatively reconstructed as the primary attachment for the FDLh in Regisaurus (state ∼3, level II′ inference; parafibula is unknown in non-cynodont therapsids).

Character 56. Baurioids including Regisaurus can be explicitly coded for state 1 here as they possess a distinct calcaneal tuber (state 1, level II inference).

Character 61. Following the same logic as outlined above for Oudenodon and Lycaenops, the attachment in Regisaurus is assumed to be centered on the femur (approximating state 1, level II′ inference).

Character 62. In the absence of any positive osteological evidence, the plesiomorphic state is reconstructed here for Regisaurus (state 0, level II′ inference).

Character 64. Following the same logic as outlined above for Oudenodon and Lycaenops, the EDBh in Regisaurus is reconstructed as originating from the anterior aspect of the distal fibula (state 2, level II′ inference).

Character 66. Following the same logic as outlined above for Oudenodon and Lycaenops, the origin of the EHL in Regisaurus is reconstructed at the fibular midshaft (state 2, level II′ inference).

Character 69. An insertion of the TA distal to the ankle is ubiquitous among amniotes, with the general locus of attachment located at metatarsal I. Therocephalians already had achieved the tarsal arrangement observed in extant mammals (Schaeffer, 1941; Boonstra, 1964; Kemp, 1986; personal obsservation of BP and SAM specimens), suggesting that Regisaurus likely possessed a more mammal-like attachment of the TA, on the entocuneiform in addition to metatarsal I (state 3, level II′ inference).

Character 71. Following the same logic as outlined above for Oudenodon and Lycaenops, the PL in Regisaurus is inferred to have originated from around the proximal fibula (state 3, level II′ inference).

Character 72. Following the same logic as outlined above for Oudenodon and Lycaenops, the PB likely originated from the anterolateral proximal fibula in Regisaurus (state 3, level II′ inference).

Character 74. Following the same logic as outlined above for Oudenodon and Lycaenops, the PB of Regisaurus likely inserted on metatarsal V (state 4, level II′ inference).

Character 76. Therocephalians already had achieved the tarsal arrangement observed in extant mammals, suggesting that Regisaurus may have possessed a more mammal-like attachment of the PP on the plantar aspect of the proximal tarsals (state 3, level II′ inference).

Massetognathus pascuali

The reconstruction presented in Figure 7 is based principally on MCZ VPRA-3691, which comprised the disarticulated skeletons of several individuals and one additional individual that remains nearly fully articulated (Jenkins, 1970). Additional material of Massetognathus in the MCZ and PVL collections helped refine observations and inferences. Phalangeal morphology is principally based on Thrinaxodon liorhinus (Fernandez et al., 2013). Inferred origins and insertions are described in Table 6. Massetognathus was codified as part of Cynognathia, and reconstructed states at the node Eucynodontia ( Table S6 (Tables_S1-8.xlsx)) were used to clarify remaining uncertainties. Some character states remained ambiguous for this node, and these were resolved for Massetognathus as follows.

Figure 7.

Reconstructed muscle attachments in Massetognathus. A, pelvis and axial skeleton, in lateral and ventral views. B, femur in anterior, lateral, posterior, and medial views. C, tibia and fibula in anterior, lateral, posterior, and medial views. D, pes in dorsal (top) and ventral (bottom) views; phalanges based in Thrinaxodon. Red denotes inferred origins, blue denotes inferred insertions, and orange denotes ambiguous attachments (see text). Scale = 10 mm. See Table 6 for muscle abbreviations.

Table 6.

Reconstructed origins and insertions in Massetognathus pascuali. LOI, level of inference in the scheme of Witmer (1995).

Table 6.

Continued.

Character 2. Although the state is unambiguous for Massetognathus, the exact location on the ilium is ambiguous, given the uncertainty in precisely when the migration of the ITa to RF took place. The origin is tentatively reconstructed on the anterior margin of the ilium, but it is possible that it may have already become more ventrally situated by the appearance of Eucynodontia.

Character 12. The apomorphic anterior expansion of the cynodont iliac blade suggests enlargement and probable differentiation of the primitive iliofemoralis mass, and so two heads are conservatively reconstructed here (state 1, level II′ inference).

Character 15. Without any positive evidence for a splitting of the muscle mass, the PIFI is conservatively reconstructed as singular here (state 0, level II′ inference).

Character 18. There is no positive evidence for a splitting of the PIFE muscle mass in cynodonts, and conservatively a single head would be reconstructed. However, as explained for the non-cynodont therapsids, the relatively broad area of the inferred origin (character 19) suggests that the PIFE likely possessed some form of anteroposterior subdivision, correlating with the mammalian OBEX (anterior) and QF (posterior).

Character 27. Following the same line of reasoning as for the therapsids, an insertion on the tibia is reconstructed for Massetognathus (state 1, level II′ inference).

Character 28. The presence of the FTE is most parsimoniously inferred as a synapomorphy of Amniota (or a more inclusive clade), so only a single head is inferred for the FTE here (state 1, level II′ inference).

Character 29. As per the previous character, the FTE is inferred as present. The appearance of the ischiadic tuberosity in eucynodonts potentially signaled a topographic shift/reorganization of the crural flexor musculature (Bishop and Pierce, 2024b), whereby the FTE shifted from dorsal to the ischium to becoming more strongly associated with the ischium. To this end, an origin from near the ischiadic tuberosity is inferred here for Massetognathus (state 3, level II′ inference), occupying a position corresponding more closely to the posterior (ventral) division of crown mammals.

Character 31. Following the same line of reasoning as for the therapsids, a secondary attachment to the GL was absent in Massetognathus (state 2, level II′ inference).

Character 34. Cynodonts are inferred to have already exhibited a shift in PIFI origin (character 16, state 2), which in turn suggests a change in PUT topology from the plesiomorphic, single-headed condition (see Bishop and Pierce, 2024b). Two heads and two origins are reconstructed in here in Massetognathus (state 2, level II′ inference).

Character 35. Following the same logic as for the preceding character, two femoral insertions are reconstructed in Massetognathus (state 2, level II′ inference), corresponding to the PECN and ADDL.

Character 37. As the ADD is unambiguously reconstructed as possessing a single head (character 36), this obviates the possibility of a crocodylian-like origin. However, it is not possible to discern between the remaining potential states in the absence of osteological evidence. As the puboischiadic plate is already markedly transformed in eucynodonts, the more mammalian condition is tentatively inferred in Massetognathus (state 3, level II′ inference), reconstructing the origin at the ventral junction of the pubis and ischium.

Character 43. The greatly reduced tail of cynodonts, including Massetognathus (Jenkins, 1970), suggests a more mammalian type of origin for the CF. The monotremes' condition is reconstructed here for Massetognathus, whereby an osseous contact is still retained (state 2, level II′ inference).

Character 46. The gastrocnemii are ubiquitous among amniotes, and so the GE can be confidently inferred as present in cynodonts. The development of a distinct calcaneal tuber in cynodonts (including Massetognathus) may signal increasing differentiation into multiple heads, but crocodylians also possess a calcaneal tuber and an undivided GE. Even in well-preserved, articulated skeletons of cynodonts (including those of Massetognathus), there is no evidence for ossified fabellae, or a parafibula, which might signal subdivisions. Without any positive evidence for a splitting of the muscle mass, the GE is conservatively reconstructed as singular here (state 1, level II′ inference).

Character 48. As per character 46, a distinct PLA was not present in Massetognathus (state 0, level II′ inference).

Character 52. As per character 46, a distinct PLA was not present in Massetognathus (state 0, level II′ inference).

Character 55. The locus of the FDLh origin in amniotes is generally centered around the posterior proximal fibula; this area is conservatively reconstructed as the primary attachment for the FDLh in Massetognathus (state ∼3, level II′ inference; parafibula is unknown in basal cynodonts).

Character 61. Following the same logic as outlined above for the therapsids, the attachment in Massetognathus is assumed to be centered on the femur (approximating state 1, level II′ inference).

Character 62. In the absence of any positive osteological evidence, the plesiomorphic state is conservatively reconstructed here for Massetognathus (state 0, level II′ inference).

Character 64. Following the same logic as outlined above for the therapsids, the EDBh in Massetognathus is reconstructed as originating from the anterior aspect of the distal fibula (state 2, level II′ inference).

Character 66. Following the same logic as outlined above for the therapsids, the origin of the EHL in Massetognathus is reconstructed at the fibular midshaft (state 2, level II′ inference). Jenkins (1971, fig. 58) described a ridge on the anterior fibula of Cynognathus (or Diademodon), which may signify the attachment of this muscle in that taxon.

Character 69. An insertion of the TA distal to the ankle is ubiquitous among amniotes, with the general locus of attachment located at metatarsal I. Cynodonts already had achieved the tarsal arrangement observed in extant mammals (Bonaparte, 1963; Jenkins, 1971; De Oliveira et al., 2010; Fernandez et al., 2013; Liu et al., 2017), suggesting that Massetognathus may have possessed a more mammal-like attachment of the TA, on the entocuneiform in addition to metatarsal I (state 3, level II′ inference).

Character 71. Following the same logic as outlined above for the therapsids, the PL in Massetognathus is inferred to have originated from around the proximal fibula (state 3, level II′ inference).

Character 72. Following the same logic as outlined above for the therapsids, the PB likely originated from the anterolateral proximal fibula in Massetognathus (state 3, level II′ inference).

Character 74. Following the same logic as outlined above for the therapsids, the PB of Massetognathus likely inserted on metatarsal V (state 4, level II′ inference).

Character 76. Cynodonts already had achieved the tarsal arrangement observed in extant mammals, suggesting that Massetognathus may have possessed a more mammal-like attachment of the PP on the plantar aspect of the proximal tarsals (state 3, level II′ inference).

Megazostrodon rudnerae

The reconstruction presented in Figure 8 is based principally on the holotype, NHMUK PV M.26407, a partially articulated individual for which the postcranial elements were originally described by Jenkins and Parrington (1976). The pubis, proximal tibia, distal tibia and fibula, and a few pedal elements are missing, so those parts of the anatomy were based on the closely related morganucodontids Erythrotherium parringtoni and Eozostrodon parvus. In addition to the descriptions provided by Jenkins and Parrington (1976), information on both taxa was also sought from firsthand observation of material (SAM-PK-K359 for Erythrotherium, and various specimens in the MCZ collections for Eozostrodon). Inferred origins and insertions are described in Table 7. Megazostrodon was codified as part of Morganucodontidae, and reconstructed states at the node Mammaliaformes ( Table S7 (Tables_S1-8.xlsx)) were used to clarify remaining uncertainties. Some character states remained ambiguous for this node, and these were resolved for Megazostrodon as follows.

Figure 8.

Reconstructed muscle attachments in Megazostrodon. A, pelvis and axial skeleton, in lateral and ventrolateral views; although unknown from fossils, an epipubis (dashed outline) was possibly present in Megazostrodon. B, femur in anterior, lateral, posterior, and medial views. C, tibia and fibula in anterior, lateral, posterior, and medial views. D, pes in dorsal (top) and ventral (bottom) views. Red denotes inferred origins, blue denotes inferred insertions, and orange denotes ambiguous attachments (see text). Scale = 1 mm. See Table 7 for muscle abbreviations.

Table 7.

Reconstructed origins and insertions in Megazostrodon rudnerae. LOI, level of inference in the scheme of Witmer (1995).

Table 7.

Continued.

Table 7.

Continued.

Character 12. Following the same logic as for Massetognathus, two heads are conservatively reconstructed here for Megazostrodon (state 1, level II′ inference). The development of a longitudinal ridge and gluteal fossa on the ilium of derived probainognathian cynodonts also suggests that this part of the hindlimb had closely approached the mammalian condition, further supporting the inference of multiple heads here.

Character 15. Without any positive evidence for a splitting of the muscle mass, the PIFI is conservatively reconstructed as singular here (state 0, level II′ inference; an “iliopsoas”).

Character 18. There is no positive evidence for a splitting of the muscle mass in probainognathians, and conservatively a single head would be reconstructed. However, the inferred origin (character 19) nonetheless covers a broad area, and so in a similar fashion to Massetognathus, an anteroposterior subdivision is considered plausible (state 1, level II′ inference).

Character 27. The FTI2 is synapomorphic for Amniota (character 23), and so it can be inferred to have existed in Megazostrodon. As hypothesized by Bishop and Pierce (2024b), the persistence of a fibular ridge suggests that the switch from ILFB-dominated (TEN-dominated) to FTI2-dominated (BICF-dominated) attachment on the fibula probably occurred crownward of Mammaliaformes. Nevertheless, not that further up the stem lineage at crown Mammalia, the BICF dominates the lateral attachment. It is therefore probable that an intermediary or transitional attachment on the posterior fascia of the crus existed in Megazostrodon (state 1/2, level II′ inference; cf. Romer, 1922, plate 39).

Character 28. The presence of the FTE is most parsimoniously inferred as a synapomorphy of Amniota (or a more inclusive clade), so only a single head is inferred for the FTE here (state 1, level II′ inference).

Character 29. Following the line of reasoning outlined for Massetognathus, an origin from near the ischiadic tuberosity is inferred here for Megazostrodon (state 3, level II′ inference). The tuberosity is well developed in morganucodontids, including Megazostrodon.

Character 31. Following the same line of reasoning as for Massetognathus, a secondary attachment to the GL was absent in Megazostrodon (state 2, level II′ inference).

Character 34. Following the same logic as for Massetognathus, a shift in PIFI origin and PUT recombination are inferred to have taken place prior to Mammaliaformes. The presence of the typical “trihedral” iliac structure in morganucodontids (first appearing around Prozostrodon) supports the inference of a typically mammalian configuration, and two heads and two origins are therefore reconstructed in Megazostrodon (state 2, level II′ inference; PECN and ADDL).

Character 35. Following the same logic as for the preceding character, two femoral insertions are reconstructed in Megazostrodon (state 2, level II′ inference), corresponding to the PECN and ADDL.

Character 37. As the ADD is unambiguously reconstructed as possessing a single head (character 36), this obviates the possibility of a crocodylian-like origin. However, it is not possible to discern between the remaining potential states in the absence of osteological evidence. As the puboischiadic plate of morganucodontids is essentially mammalian in gross structure (posteriorly swept pubis, large obturator fenestra, narrow puboischiadic rami), the mammalian condition is tentatively inferred in Megazostrodon (state 3, level II′ inference), with the origin reconstructed at the ventral junction of the pubis and ischium.

Character 43. NHMUK PV M.26407 preserves at least five well-developed caudal vertebrae in articulation with the sacrum, and Jenkins and Parrington (1976) interpreted the tail of Eozostrodon to comprise at least 12 vertebrae. The tail of morganucodontids was therefore of moderate length, albeit apparently thin in relation to the rest of the animal. The monotremes' condition is reconstructed here for Megazostrodon, whereby an osseous contact is still retained (state 2, level II′ inference).

Character 46. The gastrocnemii are ubiquitous among amniotes, and so the GE can be confidently inferred as present in morganucodontids. The development of an incipient parafibula is taken to indicate the differentiation of the PLA (character 48), but it is assumed here that the SOL is an apomorphy of Theria (or a more inclusive clade, e.g., Theriiformes), thus leading to state 2 being reconstructed here (level II′ inference).

Character 52. As per character 48, the PLA is inferred as present in Megazostrodon, although the exact manner of the insertion of this muscle, distal to the plantar fascia, is not certain (states 1/2, level II′ inference).

Character 55. The locus of the FDLh origin in amniotes is generally centered around the posterior proximal fibula; this area is conservatively reconstructed as the primary attachment for the FDLh in Megazostrodon. The presence of an incipient parafibula suggests that a more monotreme-like condition existed here (state ∼2, level II′ inference).

Character 61. The locus of the EDLh origin in amniotes is generally centered around the anterior knee region, with mammals typically involving more of the fibula or tibia than the femur. Given the modifications to the fibula prior to Mammaliaformes (i.e., spatulate shape, incipient parafibula), this suggests significant reorganization of the associated soft tissues as well; consequently, the origin is reconstructed as from the anterior aspect of the fibular head (state ∼2, level II′ inference), distal to the knee.

Character 62. In the absence of any positive osteological evidence, the plesiomorphic state is reconstructed here for Megazostrodon (state 0, level II′ inference), although its close phylogenetic proximity to mammals may argue for state 1 being more likely instead.

Character 64. Plesiomorphically, the general locus of the origin for the EDBh is the ankle joint, but extant mammals exhibit an origin near the proximal fibula. The close phylogenetic proximity to mammals and the presence of an incipient parafibula in morganucodontids suggest that the mammalian configuration is the most likely configuration (state 3, level II′ inference). The muscle's proximal path is therefore reconstructed as along the anterior face of the fibula, rather than between fibula and tibia, to reflect the muscle approaching a more “mammalian” position; however, since a lateral malleolus is unknown in morganucodontids, the path probably was not yet on the extreme lateral aspect of the fibula.

Character 66. Following the same logic as outlined above for Massetognathus, the origin of the EHL in Megazostrodon is reconstructed at the fibular midshaft (state 2, level II′ inference). Extension of the origin onto the parafibula or medial tibia is apomorphic for one genus of monotreme or another (Gambaryan et al., 2002), and so it is not phylogenetically informative.

Character 69. An insertion of the TA distal to the ankle is ubiquitous among amniotes, with the general locus of attachment located at metatarsal I. Megazostrodon possessed the same tarsal arrangement observed in extant mammals, suggesting that it possessed a more mammal-like attachment of the TA, on the entocuneiform in addition to metatarsal I (state 3, level II′ inference).

Character 71. The PL can unambiguously be inferred to have originated from around the lateral knee region, but specifically where cannot be discerned. The more mammalian construction of the bones (especially fibula) around the knee joint in Megazostrodon supports a more mammalian manifestation of muscle attachment being present. To that end, an origin from the lateral proximal fibula is reconstructed here (state 3, level II′ inference), where a large fossa provides ample area for attachment.

Character 72. Following the same logic as outlined above for Massetognathus, the PB likely originated from the anterolateral proximal fibula in Megazostrodon (state 3, level II′ inference).

Character 74. Following the same logic as outlined above for Massetognathus, the PB of Megazostrodon likely inserted on metatarsal V (state 4, level II′ inference).

Character 76. Cynodonts already had achieved the tarsal arrangement observed in extant mammals, suggesting that Megazostrodon possessed a more mammal-like attachment of the PP on the plantar aspect of the proximal tarsals (state 3, level II′ inference).

Vincelestes neuquenianus

The reconstruction presented in Figure 9 is based on material belonging to several similarly sized individuals in the MACN collections, found together as multiple associated and disarticulated skeletons (Rougier, 1993). These specimens collectively comprise all pelvic and hindlimb elements except the distal tarsals, metatarsals, and phalanges, which are poorly represented in the available material. Anatomy of the distal pes was thus based on the more completely known pedes of Jeholodens jenkinsi (Ji et al., 1999), Zhangheotherium quinquecuspidens (Luo and Ji, 2005), Akidolestes cifellii (Chen and Luo, 2013), and Henkelotherium guimarotae (Krebs, 1991; Jäger et al., 2019). Inferred origins and insertions are described in Table 8. As noted above, Vincelestes was added as its own novel operational taxonomic unit to the data set of Bishop and Pierce (2024b); reconstructed states at the node Cladotheria ( Table S8 (Tables_S1-8.xlsx)) were used to clarify remaining uncertainties. Some character states remained ambiguous for this node, and these were resolved for Vincelestes as follows.

Figure 9.

Reconstructed muscle attachments in Vincelestes. A, pelvis and axial skeleton, in lateral and ventrolateral views. B, femur in anterior, lateral, posterior, and medial views. C, tibia and fibula in anterior, lateral, posterior, and medial views. D, pes in dorsal (top) and ventral (bottom) views; distal tarsals and phalanges based on Jeholodens, Zhangheotherium, Akidolestes, and Henkelotherium. Red denotes inferred origins, blue denotes inferred insertions, and orange denotes ambiguous attachments (see text). Scale = 10 mm. See Table 8 for muscle abbreviations.

Table 8.

Reconstructed origins and insertions in Vincelestes neuquenianus. LOI, level of inference in the scheme of Witmer (1995).

Table 8.

Continued.

Table 8.

Continued.

Character 1. Without any positive evidence for a splitting of the ITp into separate heads (i.e., GMAX and FCOC), only a single head is inferred here (state 0, level II′ inference).

Character 5. Unlike many other Mesozoic crown mammals, Vincelestes lacks any indication of a third trochanter or other crest on the lateral femoral shaft (state 3, level II′ inference).

Character 6. The origin of the AMB is presently unable to be resolved, and hence it is unclear whether the muscle originated more ventrally, in the plesiomorphic configuration (i.e., AMB-like), or more dorsally from the ilium, in the therian configuration (i.e., SART-like). In the absence of osteological evidence, it is currently not possible to discriminate between the two states.

Character 11. Although optimization here reconstructs the patella as ancestrally present for cladotherians, previous work has suggested that ossified patellae evolved independently in marsupials and placentals (Samuels et al., 2017). A patella has not yet been identified in Vincelestes, although this may well be due to the disarticulated nature of most of the available material. A patella is not present in the closely related cladotherian Henkelotherium guimarotae, which is represented by articulated material (Jäger et al., 2019).

Character 12. Assuming that a twofold subdivision of the IF mass was a precursor to a threefold subdivision on the line to therians, this precludes state 0. Without any positive evidence either way, only two gluteal heads are reconstructed here in Vincelestes (state 1, level II′ inference). This captures the two main heads that are consistently present in extant mammals; other heads such as the pyriformis are much smaller and tend to be positioned closer to the hip joint, thus having relatively low functional significance.

Character 15. Without any positive evidence for a splitting of the muscle mass, the PIFI is conservatively reconstructed as singular here (state 0, level II′ inference; an “iliopsoas”).

Character 22. Parsimony recovers the ILFB as absent for Cladotheria, which is incongruous with the reconstruction for character 21 and its presence in some therians. An indirect, soft-tissue attachment to the lateral crural fascia is inferred here (state 2, level III′ inference), which is the condition observed in therians when the muscle is present.

Character 36. The puboischiadic plate of Vincelestes is notably large in relation to the ilium, compared to that typically observed in therians, and this would provide ample room for the origin of two separate adductor masses. Both short (ADDB) and large (ADDM) heads are reconstructed here (state 1, level II′ inference).

Character 39. As discussed by Bishop and Pierce (2024b), it is not currently possible to discern whether the single-headed condition in monotremes is plesiomorphic or apomorphic. A single head is conservatively assumed here (state 0, level II′ inference), although in functional terms, the presence of a second head (i.e., the GEM) would be relatively small and subsidiary to the larger muscle belly (i.e., the OBIN).

Character 43. Vincelestes possesses a very sizeable tail for a mammal (≫30 caudal vertebrae; Rougier, 1993), and its proximal caudal vertebrae bear well-developed transverse processes. An origin similar to that observed in monotremes is tentatively assumed here, in which the CF retains an osseous contact (state 2, level II′ inference).

Character 46. The gastrocnemii are ubiquitous among amniotes, and so the GE can be confidently inferred as present in Vincelestes. The development of an incipient parafibula in more basal mammals is taken to indicate the differentiation of the PLA prior to the origin of Cladotheria, but the SOL is assumed to be an apomorphy of Theria; state 2 is reconstructed here (level II′ inference).

Character 48. Irrespective of state 1 or 2, the locus of attachment is somewhere around the proximal fibula. It is unclear the extent to which Vincelestes possessed a parafibula, although the morphology of the structure preserved at the proximal end in MACN-N 38 is comparable to the structure observed in other taxa with such a process (e.g., Megazostrodon, basal therians), and so state 1 is tentatively inferred (level II′ inference).

Character 49. As per character 46, the SOL is presumed to be absent in Vincelestes (state 0, level II′ inference).

Character 52. As per character 48, the PLA is inferred to be present in Vincelestes, although the exact manner of the insertion of this muscle, distal to the plantar fascia, is not certain (states 1/2, level II′ inference).

Character 53. As per character 46, the SOL is presumed to be absent in Vincelestes (state 0, level II′ inference).

Character 55. The locus of the FDLh origin in amniotes is generally centered around the posterior proximal fibula, and this area is conservatively interpreted as the primary attachment for the FDLh in Vincelestes. In lieu of definitive evidence of a parafibula in Vincelestes, the origin is reconstructed as from the posterior aspect of the fibular head (state 3, level II′ inference).

Character 59. It is unclear if the FHL of therians is homologous with the FHL observed in crocodylians and Sphenodon, and indeed the “therian FHL” may be a therian apomorphy. If it was present in Vincelestes, it would have presumably taken origin from the posteroproximal tibia (state 2).

Character 60. Following the same reasoning as the previous character, if a FHL were present in Vincelestes, it would have presumably inserted in a therian-like fashion (state 2).

Character 61. The locus of the EDLh origin in amniotes is generally centered around the anterior knee region, with mammals typically involving more of the fibula or tibia than the femur. Given the modifications to the fibula prior to Mammaliaformes (i.e., spatulate shape, incipient parafibula), this suggests significant reorganization of the associated soft tissues as well, and the origin is tentatively reconstructed here as from the anterior aspect of the fibular head (state ∼2, level II′ inference).

Character 63. It remains uncertain as to when any “extra peroneal heads” differentiated on the line to therians, and hence it is currently ambiguous whether the EDBh was accompanied by other bellies in Vincelestes.

Character 66. The EHL is ubiquitous among amniotes and can be confidently inferred as present in Vincelestes. Its locus of origin is centered around the anterior fibula, and in the absence of any positive evidence, a middle-of-the-road approach is taken and the origin reconstructed at the fibular midshaft (state 2, level II′ inference). Extension of the origin onto the parafibula or medial tibia is apomorphic for one genus of monotreme or another, and so it is not considered here to be phylogenetically informative.

Character 71. The PL can unambiguously be inferred to have originated from around the lateral knee region, but specifically where cannot be discerned. The more mammalian construction of the bones (especially the lightly constructed fibula) around the knee joint in Vincelestes supports a more mammalian manifestation of muscle attachment being present. The origin is here reconstructed as from the lateral proximal fibula (state 3, level II′ inference), where a large fossa provides ample area for attachment.

Character 73. The calcaneum of Vincelestes possesses a well-developed lateral corner to the peroneal shelf and a well-formed tuber, which are elsewhere inferred as probable osteological indicators of states 1 and 3, respectively. However, as a crown mammal, it is likely to have possessed an insertion more on the ventral aspect of the medial pes (Bishop and Pierce, 2024b). As such, although the PL may have gained attachment to the peroneal shelf, a terminal insertion on the plantar aspect of metatarsal 1 is inferred here (state 4, level II′ inference).

Character 74. The existence of two peroneal heads is supported here (character 70), with the PB inserting in a mammal-like fashion somewhere along pedal ray V. Vincelestes possesses a pronounced lateral projection on the base of metatarsal V (Rougier, 1993), and the PB is inferred to have inserted on this process (state 4, level II′ inference).

Character 75. The exact origin of the PP in Vincelestes is uncertain, but the locus is inferred to be the posterior aspect of the crus, deep to the FDLh but superficial to the POP.

Character 77. The origin for the POP in Vincelestes is uncertain, although the general locus of attachment is around the proximal fibula, and this is conservatively reconstructed as the principal site of origin (∼state 0, level II′ inference), with limited, if any, attachment to the femur.

Comparison to Prior Studies

The reconstructions presented here offer a new, more comprehensive account of hindlimb muscular anatomy across key stages of (pretherian) synapsid evolution. For some of the focal taxa considered here, this represents the first extensive interpretation of the fossil evidence, whereas in other taxa, it represents a revised update upon older works (Romer, 1922; Colbert, 1948; Kemp, 1978; DeFauw, 1986). Short of a muscle-by-muscle, study-by-study comparison to the interpretations of previous authors, some of which were already noted above, a few overarching points may be made here to provide a broader context for evaluating the merit of the present reconstructions.

There are numerous points of minor difference in comparison to prior studies, but the reconstructions provided here are nonetheless broadly consilient with previous interpretations of the general locus of attachment of most hip muscles. More importantly, however, the present study provides the first detailed interpretation of all muscles crossing the knee and ankle joints, in addition to those crossing the hip. It therefore offers an important basis for interpreting crural and pedal anatomy in fossil taxa. Whereas many prior studies have interpreted fossil hindlimb osteology with respect to musculature, the majority of such interpretations were restricted to hip muscles (e.g., Parrington, 1961; Boonstra, 1964; Jenkins, 1971; Jenkins and Parrington, 1976; Kemp, 1978, 1980a,b; King, 1981, 1985; Sullivan et al., 2013; Guignard et al., 2018, 2019). This is probably due, in no small part, to the anatomical focus of the seminal works of Gregory and Camp (1918) and Romer (1922), the latter of which serving as the sine qua non point of reference for almost all subsequent studies over the following 100 years. Even then, subsequent studies often did not reconstruct every hip muscle, instead tending to focus on those of greater evolutionary significance, especially the IF, PIFI, PIFE, and CF, i.e., those which underwent major reorganization on the line to mammals. Several studies have also briefly considered biarticular hip muscles that crossed the knee, those comprising the triceps femoris and crural flexor (“hamstring”) groups, but rarely have these muscles' crural insertions been given the same level of attention as their pelvic origins (Gregory and Camp, 1918; Romer, 1922; Colbert, 1948; DeFauw, 1986; Ray and Chinsamy, 2003; Ray, 2006). Only Walter (1988) provided any detailed interpretation of muscles originating distal to the knee.

Reconstructing only a subset of the full complement of limb muscles can only ever offer an incomplete picture of anatomy in a given extinct species and its evolution through time. More critically, attempting to do so runs the risk of failing to account for the attachments of other muscles that are not considered, yet which “competed” for attachment space on the available bony surfaces. Thus, prior studies have at times interpreted the IF as originating from the entire iliac blade, affording insufficient space for the ILFB or IT, or they have reconstructed the PIFE as originating from the entire ischium, affording insufficient space for the ADD, FTI, or PIT (e.g., Kemp, 1980a,b; King, 1981; Guignard et al., 2018). The simplistic and erroneous reconstructions that result can hinder assessments of muscle functional evolution, especially if the unreconstructed muscles were an important functional complement (as agonist or antagonist) to the muscles that were reconstructed.

One additional point worth noting is that hypotheses of limb muscle homology across tetrapods have changed considerably over the past 100 years, stemming from advances in anatomical and embryological comparative data sets (see Diogo and Molnar, 2014; Diogo et al., 2016). Several points of difference between the present set of reconstructions and those of certain prior studies may be attributed to the use of now-outdated homology schemes by prior studies. For example, King (1985) and DeFauw (1986) reconstructed the ILFB as originating from the ischium in various dicynodonts, assuming homology of this muscle and the mammalian BICF. The ILFB and BICF are now understood to be nonhomologous, and as such, the former is consistently reconstructed here as originating dorsally from or near the posterior ilium, and the latter (or its homologue, the FTI2) is consistently reconstructed as originating ventrally from the posterior ischium. Consideration of this issue is essential to undertaking an evenhanded appraisal of historical interpretations, especially if these are to form the basis for new investigations.

Uncertainty in Reconstructions

While the present set of reconstructions is more comprehensive than prior attempts, and the current reconstructions are based on the largest data set for extinct and extant anatomy thus far assembled (Bishop and Pierce, 2024b), they are not infallible. Various aspects come with a level of uncertainty, resulting from ambiguity in the interpretation of available evidence, and they may be subject to re-evaluation in light of new fossil or developmental data. Many interpretations are only modestly supported, qualifying as level II′ inferences in the credibility scheme of Witmer (1995). Contrasting reconstructions may have important consequences for higher-level functional interpretations such as locomotor performance and behavior.

In addition to a given muscle's presence or absence, and the general locus of its origins and insertions with respect to the underlying skeleton, the muscle maps presented here also outline interpreted areas of attachment for individual muscles. Given the paucity of direct osteological evidence for precisely delimiting attachment area in most muscles (especially those with fleshy attachments), reconstructing these areas remains a subjective process and therefore carries a higher level of uncertainty compared to other aspects of a reconstruction (e.g., presence or absence). Such uncertainty is likely to be more prevalent in extinct taxa that are more phylogenetically distant or morphologically disparate from their nearest extant relatives. As noted by Bryant and Seymour (1990), error in reconstructed muscle attachment areas may lead to error in estimates of muscle size or strength (i.e., force-generating capacity). However, even in extant species where attachment areas have been quantitatively measured, the correlation between attachment area and muscle strength can be of variable quality and can vary considerably across the specific muscles and taxa considered (Fahn-Lai et al., 2020; Bates et al., 2021; Cuff et al., 2023). Errors in estimated muscle attachment area may therefore be secondary to other sources of error or variation in their possible consequences for downstream analyses. Furthermore, growing numbers of studies have developed alternative approaches to estimating muscle strength in extinct species, which do not require quantitative data on attachment areas, yet which remain grounded in empirical data sets derived from extant taxa (e.g., Sellers et al., 2013; Bates and Falkingham, 2018; Bishop et al., 2021a,b). Thus, imprecision in reconstructed muscle attachment areas may not actually matter in some situations.

Beyond estimation of muscle strength, reconstructions of attachment areas can potentially influence functional analyses by determining the spatial distribution of muscular forces in a biomechanical model. The centroid of a given attachment area is frequently used to guide the creation of muscle lines of action in musculoskeletal (rigid body) models and may therefore influence a muscle's length or its moment arm(s) with respect to joint angle (Brassey et al., 2017; Bishop et al., 2018, 2021a; Brocklehurst et al., 2022). Muscle attachment areas, or their centroids, also dictate how muscle forces are applied to bony structures in finite element models (Sellers et al., 2017; Bishop et al., 2018; Cost et al., 2020). The extent to which error in reconstructed attachment areas may significantly affect downstream analyses and interpretations in these contexts is seldom investigated (Brassey et al., 2017), although it might be expected that approaches based on area centroids should be less sensitive than those that are not, since centroids will be more robust to finer-scale variation in attachment boundaries. Considering the great diversity in musculoskeletal anatomy and function encountered in Synapsida (let alone broader taxonomic groups), and the diversity of approaches used to study musculoskeletal function, it is difficult to draw overarching generalizations about how uncertainty in muscle attachment reconstruction may affect higher-level inferences. To that end, future studies that draw upon the present set of reconstructions should carefully consider the aspects (if any) of uncertainty that are most critical to the question at hand and conduct appropriate sensitivity analyses to evaluate the consequences of alternative reconstructions (Bishop et al., 2021a).