Wing Musculature Reconstruction in Extinct Flightless Auks (Pinguinus and Mancalla) Reveals Incomplete Convergence with Penguins (Spheniscidae) Due to Differing Ancestral States

Synopsis Despite longstanding interest in convergent evolution, factors that result in deviations from fully convergent phenotypes remain poorly understood. In birds, the evolution of flightless wing-propelled diving has emerged as a classic example of convergence, having arisen in disparate lineages including penguins (Sphenisciformes) and auks (Pan-Alcidae, Charadriiformes). Nevertheless, little is known about the functional anatomy of the wings of flightless auks because all such taxa are extinct, and their morphology is almost exclusively represented by skeletal remains. Here, in order to re-evaluate the extent of evolutionary convergence among flightless wing-propelled divers, wing muscles and ligaments were reconstructed in two extinct flightless auks, representing independent transitions to flightlessness: Pinguinus impennis (a crown-group alcid), and Mancalla (a stem-group alcid). Extensive anatomical data were gathered from dissections of 12 species of extant charadriiforms and 4 aequornithine waterbirds including a penguin. The results suggest that the wings of both flightless auk taxa were characterized by an increased mechanical advantage of wing elevator/retractor muscles, and decreased mobility of distal wing joints, both of which are likely advantageous for wing-propelled diving and parallel similar functional specializations in penguins. However, the conformations of individual muscles and ligaments underlying these specializations differ markedly between penguins and flightless auks, instead resembling those in each respective group’s close relatives. Thus, the wings of these flightless wing-propelled divers can be described as convergent as overall functional units, but are incompletely convergent at lower levels of anatomical organization—a result of retaining differing conditions from each group’s respective volant ancestors. Detailed investigations such as this one may indicate that, even in the face of similar functional demands, courses of phenotypic evolution are dictated to an important degree by ancestral starting points.


29
Convergent evolution, in which distantly-related lineages acquire similar traits, has been 30 regarded as evidence for the predictability of organismal evolution under natural selection 31 (e.g., Conway Morris 2003, 2010Melville et al. 2006;Mahler et al. 2013). Convergence 32 may arise as a result of a tight relationship between phenotype and functional performance, 33 and/or evolutionary constraints or biases inherent to certain organismal designs that result in 34 a limitation of possible phenotypic solutions (Wake 1991;Losos 2011;Wake et al. 2011). 35 These factors often operate simultaneously, and may lead to idiosyncratic outcomes because 36 of differences in ancestral conditions and/or evolvability between lineages (historical 37 contingency; e.g., Gould 2002; Agrawal 2017;Blount et al. 2018). Recent studies have 38 demonstrated that idiosyncrasies among lineages occupying similar niches, termed 39 "incomplete" convergence (sensu Herrel et al. 2004), might be more prevalent than 40 previously recognized (Losos 2010;Moen et al. 2016;Hulsey et al. 2019). As such, close 41 examination into the nature of apparently convergent phenotypes and their ancestral 42 conditions is required to fully comprehend the various evolutionary processes underlying 43 convergence. 44 The evolution of avian wing-propelled diving provides a classic example of 45 convergent evolution. Wing-propelled diving describes a mode of underwater locomotion 46 whereby birds propel themselves by flapping their forelimbs (aquatic flight ;Townsend 1909;47 12 study. Subfossil bones of Pinguinus impennis from the collections of the Museum of 249 Zoology, University of Cambridge, Cambridge, UK (UMZC 187.d and 187.G) were the 250 primary source of osteological data for the musculature reconstruction for this species. 251 Thousands of fossil specimens of Mancallinae are available in museum collections, 252 from which several species of Mancalla have been described (e.g., Chandler 1990;Smith 253 2011). Nevertheless, no single specimen represents a sufficient component of the pectoral 254 girdle and wing skeleton to enable reconstruction of the complete musculature of the wing. 255 This situation necessitated that the reconstructions be based on observations of multiple 256 specimens. Several associated, partial skeletons were available in the extensive collections of 257 fossil birds at the Natural History Museum of Los Angeles County, Los Angeles, and the San 258 Diego Museum of Natural History, San Diego (both California, USA). These constituted the 259 primary basis of the reconstructions for Mancalla. In addition, other well-preserved 260 specimens were also examined to complement observations on these associated skeletons 261 (Table 2). Many Mancalla specimens could not be identified to species level due to a lack of 262 diagnostic features, and they may represent multiple species. Nevertheless, only negligible 263 qualitative variation was observed in the relative positions of osteological correlates. Thus, 264 these specimens were collectively treated as representatives of a single taxon (Mancalla) for 265 the purpose of wing musculature reconstruction. Although this is admittedly a coarse 266 assumption that may overlook potential interspecific variation, this approach was necessary 267 in the light of a lack of complete skeletons. Some of the specimens had originally been 268 identified as belonging to species that were considered invalid by Smith (2011), but no 269 attempt was made to re-identify them to species level, apart from confirming their assignment 270 to Mancalla. 271 There was no complete sternum for Mancallinae available for the present study, nor 272 have any been reported in the literature. As a result, the sternal morphology of Mancalla 273 13 needed to be reconstructed from multiple specimens. This reconstruction was accomplished 274 by photographic collage of three well-preserved partial sterna; photographs of two specimens 275 (LACM 2180 and SDSNH 77399), taken in lateral and ventral views, were overlaid onto 276 photographs of another specimen (SDSNH 26242), with the former ones rescaled such that 277 the outlines of their preserved portions matched those of the latter as closely as possible, 278 while retaining their original aspect ratios. This procedure may have introduced some 279 inaccuracies in scaling into the reconstruction of this portion of the skeleton, as the specimens 280 involved differed distinctly in size. 281 Osteological correlates on the bones of the extinct species were identified by 282 comparison with those in extant species, based on their shape, nature (e.g., tubercles, scars, 283 lines), and positions relative to other landmarks. In most cases, the presence of muscles and 284 ligaments could be inferred by "Level I" inferences of the Extant Phylogenetic Bracket 285 framework (Witmer 1995(Witmer , 1997. That is, the presence of a muscle/ligament in an extinct 286 species was inferred based on the presence of the corresponding osteological correlate in that 287 species and the conserved relationship between the soft parts and the osteological correlate in 288 at least two extant species that phylogenetically "bracket" the extinct species. In some cases, 289 however, only weaker inferences could be made, for which specific notes are given below. 290 For Pinguinus impennis, a dried partial skeleton with remnants of the elbow and 291 forearm musculature (NHMUK 1972.1.156) became available after the reconstruction based 292 on osteological correlates was complete. The reconstructed musculature was subsequently 293 compared with this desiccated specimen in order to verify the validity of the reconstruction 294 based only on osteological correlates. 295 296

Musculature in extant birds 308
Among the charadriiform birds examined, most wing ligaments and muscles were observed 309 in generally consistent positions. In many cases, the attachments of ligaments and tendons 310 (indirect attachments of muscles) corresponded to distinct tubercles or scars, which could be 311 easily delineated. In contrast, the margins of fleshy (direct) attachments could not be clearly 312 discerned unless delineated by intermuscular lines or other osteological landmarks, as pointed 313 out previously (Bryant and Seymour 1990). Descriptions of major wing ligaments and 314 muscles are given below, as well as illustrations of the overall musculature in a representative 315 taxon (Alca; Figs. 2, 3), and osteological correlates in selected taxa (Catharacta,Alca,316 Spheniscus,Pluvialis,Scolopax,Larus schistisagus,Cerorhinca,Cepphus,317 Synthliboramphus,Uria,Gavia,Ardenna,. Results for 318 Larus crassirostris, Fratercula, and Calonectris were mostly similar to those of Larus 319 schistisagus, Cerorhinca, and Ardenna, respectively. 320 321

Ligaments of the shoulder 322
The lig. acrocoracohumerale is a prominent ligament connecting the proximal end of the 323 humerus to the processus acrocoracoideus of the coracoid (Fig. 3). Its origin on the coracoid 324 is marked by a broad scar (impressio lig. acrocoracohumeralis) on the dorsolateral margin of 325 the processus acrocoracoideus, typically between the facies articularis humeralis and the omal 326 end of the coracoid (Figs. 4,5,10,11). Its humeral insertion lies on the ventral margin of the 327 sulcus transversus on the cranial aspect of the proximal humerus (Figs. 6,7,12,13). In 328 Spheniscus, the caudodorsal part of this ligament is somewhat differentiated, and could be 329 termed the lig. coracohumerale dorsale; its origin extends onto the dorsal margin of the 330 glenoid cavity, and its insertion is on the craniodistal margin of the sulcus transversus, 331 adjacent to the typical insertion of the lig. acrocoracohumerale (Figs. 19,21). 332 In most taxa examined, a thick, distinct ligament or retinaculum bridges between the 333 lateral margin of the collum scapulae and the caudodistal margin of the caput humeri, 334 providing an origin for the m. scapulotriceps (Fig. 3). This ligament is apparently not 335 formally named in Baumel and Raikow (1993). Here, this ligament is tentatively referred to 336 as the retinaculum originis m. scapulotricipitis. The scapular attachment of the retinaculum 337 originis m. scapulotricipitis is marked by a tubercle on the lateroventral aspect of the collum 338 scapulae (Figs. 4,5,10,11). The retinaculum is closely associated with the caudal part of the 339 joint capsule (plica synovialis transversa; below). The humeral end of this retinaculum is 340 attached to the caudodistal and ventral margins of the caput humeri (Figs. 6,7,12,13). This 341 retinaculum is absent in Spheniscus, where the m. scapulotriceps arises directly from the 342 scapula (see below). 343 The caudodorsal side of the shoulder joint capsule is sometimes developed as a 344 distinct ligament that spans between the caudal margin of the glenoid cavity and the 345 caudodistal margin of the caput humeri. This ligament is tentatively named the plica 346 the wrist, and ends as a thin aponeurosis on the dorsal surface of the proximal 570 carpometacarpus. In most taxa examined (except Gavia), the attachment scar is quite 571 indistinct, but extends distocaudally from the distal end of the attachment of the lig. 572 ulnocarpo-metacarpale dorsale (Figs. 9,17). In Gavia, the attachment of this ligament is 573 further apart distally from that of the latter ligament, and is marked by a distinct scar (Figs. 574 S27,S28). 575 The lig. obliquum alulae is a distinct ligament on the alula, originating from the 576 distocranial slope of the processus extensorius of the carpometacarpus and inserting on the 577 cranioventral margin of the proximal end of the alular phalanx. 578 The lig. collaterale caudale (of artc. metacarpophalangealis alulae) lies deep within 579 the alular articulation, connecting the caudal margins of the facies articularis alularis of the 580 carpometacarpus and the proximal end of the alular phalanx. 581 The lig. collaterale ventrale (of artc. metacarpophalangealis digiti majoris) consists of 582 two distinct parts on the ventral side of the joint. Both the cranial and caudal parts arise from 583 the ventral side of the distal end of the carpometacarpus, where the attachments are marked 584 by two distinct tubercles in Alcidae. The cranial part ends on the proximal end of the phalanx 585 (or slightly offset from the proximal articular surface in Alcidae), whereas the caudal part 586 ends on the caudal margin of the ventral surface of the proximal phalanx (Figs. 9,17). 587 The lig. collaterale caudale (of artc. metacarpophalangealis digiti majoris) is present 588 on the dorsocaudal aspect of the joint between the carpometacarpus and the proximal phalanx 589 of the major digit. Its origin is marked by a tubercle which is slightly offset from the distal 590 end of the carpometacarpus (near the level of the proximal margin of the symphysis 591 metacarpalis distalis) and lies cranial to the sulcus interosseus. The insertion is on the dorsal 592 part of the craniodorsal margin of the proximal articular surface of the phalanx (Figs. 9,17). 593 26 The lig. obliquum intra-articulare (of artc. metacarpophalangealis digiti majoris) lies 594 deep within the joint between the carpometacarpus and the proximal phalanx of the major 595 digit. It originates from the groove between the two articular surfaces of the distal end of the 596 carpometacarpus, and inserts on the caudal margin of the proximal articular surface of the 597 phalanx (Figs. 9,17). 598 The ligg. collaterale ventrale et dorsale (of artc. metacarpophalangealis digiti minoris) 599 appear to be present in most taxa examined, but they are usually not quite differentiated from 600 the articular capsule, and their attachment sites on the bones are hardly discernible. The lig. 601 interosseum (of artc. interphalangealis lateralis) connects nearly the entire cranial margin of 602 the minor digit to the caudal margin of the proximal phalanx of the major digit. 603 604

Accessory ligaments 605
The propatagium is spanned by a ligamental complex which typically consists of several 606 interconnected ligamentous bands (Fig. 2). Following Baumel and Raikow (1993), the long 607 ligamentous band forming the cranial edge of the propatagium is referred to as the lig. 608 propatagiale, whereas the caudal band running along the humerus and inserting on the 609 proximal forearm is referred to as the lig. limitans cubiti. In most taxa examined (except in 610 Spheniscus, where these ligaments are undifferentiated), these two ligaments largely share the 611 same origin. 612 In most taxa examined, the ligg. propatagiale et limitans cubiti together arise as the m. 613 deltoideus pars propatagialis (and partly as the m. pectoralis pars propatagialis; see below). 614 These are proximally anchored to the tip of the crista deltopectoralis of the humerus (Figs. 6,615 7,12,13). In Larus and Catharacta, the ligaments arise separately from the distally 616 bifurcated belly of the m. deltoideus pars propatagialis; the lig. propatagiale is further 617 bifurcated at its proximal end, with the caudal branch anchored to the crista deltopectoralis. 618 Typically, the middle part of the lig. propatagiale is flared and partly bifurcated, and around 619 the flexion of the propatagium the cranialmost part is thickened and consists of elastic fibers 620 (the so-called pars elastica). In Gavia, the pars elastica is rather enlarged, and the ligament 621 consists almost entirely of elastic fibers except near the proximal and distal ends. In 622 Procellariidae, the distal part of the ligament is largely bifurcated, and these divisions merge 623 with each other near the wrist joint. In all cases, the lig. propatagiale passes the cranial edge 624 of the wrist joint along the thickened cranioventral margin of the distal radius, where the 625 ligament hosts a sesamoid (os prominens) in Procellariidae. The ligament inserts on the 626 proximoventral margin of the processus extensorius of the carpometacarpus and the ventral 627 margin of the proximal end of the alular phalanx (Figs. S4, S6, S18, S22), but the attachment 628 sites on the bones are often hardly discernible. 629 Typically, two short branches (hereafter, the ventral and dorsal branches) arise around 630 the pars elastica of the lig. propatagiale, inserting on the dorsal and ventral sides of the 631 proximal forearm. The ventral branch is a thin ligament, and ends on the ventral fascia of the 632 proximal forearm (aponeurosis ventralis antebrachii). In Gavia, the ventral branch is also 633 anchored to the ventral surface of the belly of m. extensor carpi radialis (see below). Among 634 the taxa examined, the conformation of the dorsal branch is rather variable. It is usually 635 broad, lying on the most superficial layer of the dorsal surface of the forearm musculature. In 636 Gavia and Procellariidae, the dorsal branch seems to be merged with the lig. limitans cubiti, 637 and together these are attached to the dorsal fascia of the forearm (aponeurosis dorsalis 638 antebrachii); in turn, the fascia is anchored to the dorsal aspect of the proximal ulna with an 639 elongated scar. In Procellariidae, the dorsal branch is also attached to a sigmoidal sesamoid 640 within the cranial side of the elbow joint, which has a ligamentous connection with the tip of 641 the processus supracondylaris dorsalis of the humerus. In Charadriiformes, the dorsal branch 642 of the lig. propatagiale merges either with the aponeurosis dorsalis antebrachii, with the lig. 643 limitans cubiti, or with both of these. Either the dorsal branch of the lig. propatagiale or the 644 lig. limitans cubiti is attached on the dorsal surface of the proximal ulna, along with the lig. 645 dorsale cubiti (see above for the relative positions between the attachment sites). From the 646 attachment scars alone, it is generally impossible to discern which of the dorsal branch of the 647 lig. propatagiale or the lig. limitans cubiti is attached on the ulna. 648 The lig. limitans cubiti runs along the cranial margin of the humerus, caudal to the lig. 649 propatagiale (Fig. 2). In Charadriiformes, it is more or less distinct from, and lies deep 650 (ventral) to, the dorsal branch of the lig. propatagiale. In Alcidae, the proximal part of the lig. 651 limitans cubiti is loosely connected to the cranial margin of the humerus, where a blunt, 652 elongated ridge is present in some taxa (e.g., Uria, Synthliboramphus; Figs. S17, S18, S21, 653 S22). Therefore, the ridge is considered to be an osteological correlate indicating strong 654 attachment of the ligament to the humerus. As described above, it merges with the 655 aponeurosis dorsalis antebrachii or ends on the ulna. 656 In Spheniscus, there is virtually no distinction between the ligg. propatagiale et 657 limitans cubiti, and this single ligament is attached along the entire cranial margin of the 658 crista deltopectoralis of the humerus (Figs. 20,21). It then extends along the cranioventral 659 margin of the radius up to its insertion on the processus extensorius of the carpometacarpus 660 (Figs. 22,23). 661 The lig. humerocarpale is a long, broad ligamentous band on the superficial layer of 662 the ventral side of the forearm, connecting the distal humerus and the ulnare (Fig. 2). It arises 663 from the caudodistal-most of two distinct pits on the ventral surface of the epicondylus 664 ventralis (the other being for the m. pronator profundus; see below) (Figs. 6,7,12,13). It 665 ends on a distinct tubercle on the proximocaudal aspect of the ulnare, which lies ventral to the 666 attachment of the lig. ulno-ulnocarpale proximale (see above) (Figs. 8,9,14,15). In 667 Cepphus, the ligament is also attached on the base of the processus pisiformis of the 668 29 carpometacarpus (Fig. S14). The m. flexor digitorum superficialis arises from the deep 669 surface of this ligament (see below). In Spheniscus, the ligament arises from the ventral 670 surface of the caudodistal extension of the epicondylus ventralis (where no distinct pit is 671 discernible), proximocranial to the origin of the m. flexor carpi ulnaris (Fig. 21). The 672 ligament then becomes the m. flexor digitorum superficialis (which is entirely 673 tendinous/ligamentous; see below) without attaching to the ulnare. 674 The retinaculum m. extensoris metacarpi ulnaris is a short retinaculum that anchors 675 the proximal belly of the m. extensor carpi ulnaris to the dorsal aspect of the proximal ulna 676 (this inconsistency in terminology is as per Baumel and Raikow [1993] and Vanden Berge 677 and Zweers [1993]). The retinaculum lies deep to the lig. limitans cubiti, and also to the lig. 678 dorsale cubiti when the latter ligament is present. The ulnar attachment of the retinaculum is 679 often common with these ligaments, thus is not distinctly discernible on the bone (but see 680 Figs. 23, S4, S14, S22). 681

682
Wing muscles 683 The m. rhomboideus superficialis is a thin, sheet-like muscle connecting the scapula with the 684 vertebral column (Fig. 3). It lies deep to the m. latissimus dorsi cranialis and superficial to the 685 m. rhomboideus profundus. The muscle arises as a thin fleshy sheet from the processus 686 spinosi of several consecutive vertebrae (exact positions vary, but typically from the 687 caudalmost one or two cervical and the cranialmost few thoracic vertebrae), and ends fleshily 688 on the cranial part of the dorsal margin of the medial side of the scapular blade (Figs. 5,11). 689 In Gavia and Procellariidae, unlike in Charadriiformes, the insertion extends cranially to the 690 medial aspect of some acromial ligaments (e.g., the lig. acromioclaviculare) and associated 691 membranes (Figs. S24,S30). 692 30 The m. rhomboideus profundus is another muscle connecting the scapula with the 693 vertebral column (Fig. 3). The muscle lies slightly caudally and deep to the m. rhomboideus 694 superficialis, by which it is largely overlain. The muscle arises from the processus spinosi of 695 several consecutive vertebrae (typically from the cranialmost to caudalmost thoracic 696 vertebrae) with a partly aponeurotic origin, and ends fleshily on a broad area on the caudal 697 part of the medial surface of the scapular blade, ventral to the attachment of the m. 698 rhomboideus superficialis and dorsal to those of the mm. serratia (Figs. 5,11). 699 Three more or less distinct muscles connect the scapula with the rib cage: the m. 700 serratus superficialis pars cranialis, m. serratus superficialis pars caudalis, and m. serratus 701 profundus. All of these muscles arise as partly separate aponeuroses from the lateral surfaces 702 of some vertebral ribs; the first two typically arise from the dorsal margin of the processus 703 uncinatus, whereas the last one arises from the facies lateralis of the rib body. Typical origins 704 are the first two true (complete) ribs for the m. serratus superficialis pars cranialis, 3rd to 6th 705 true ribs for the pars caudalis, and the last floating (incomplete) and the first few true ribs for 706 the m. serratus profundus. The m. serratus superficialis pars cranialis ends as a thin but 707 distinct aponeurosis on the margo ventralis of the scapular blade, between the two heads of 708 the m. subscapularis (see below), marked by a sharp ridge. The m. serratus superficialis pars 709 caudalis ends fleshily on the medial surface of the scapula around its caudal tip. The m. 710 serratus profundus ends fleshily on the ventral area of the scapular blade, just cranial to the 711 attachment of the previous muscle (Figs. 5,11). The attachment sites of the last two muscles 712 are hardly delineated on the bones. Another muscle, the m. serratus superficialis pars 713 metapatagialis, was confirmed in most taxa examined, with the exception of Spheniscus. It 714 arises as aponeuroses from the facies laterales and/or processus uncinati of a few caudal 715 vertebral ribs, and ends fleshily on the dermis deep to the humeral feather tract. 716 31 The m. scapulohumeralis cranialis is a small muscle lying deep on the caudal aspect 717 of the shoulder joint (Fig. 3). The muscle originates fleshily from the ventrolateral aspect of 718 the collum scapulae, just caudoventral to the caudal tip of the facies articularis humeralis and 719 slightly cranioventral to the attachment of the retinaculum originis m. scapulotricipitis (Figs. 720 5,11); this area is marked by a slight depression in some alcids, e.g., Uria and 721 Synthliboramphus. The muscle ends fleshily on a restricted area within the fossa tricipitalis of 722 the humerus, just distal to the crus dorsale fossae, in the incision of the head of the m. 723 humerotriceps (see below) (Figs. 7,13). In many charadriiform taxa (Larus,Catharacta,and 724 Alcidae), the humeral attachment is marked by a slightly elevated relief, whose margins are 725 sometimes indistinct. This muscle is not present in Spheniscus (see also Schreiweis 1982). 726 The m. scapulohumeralis caudalis is a bulky muscle lying on the caudal aspect of the 727 shoulder joint (Fig. 3). It arises fleshily from most of the lateral surface of the scapula that is 728 unoccupied by other attachment sites (the origin is especially large in Spheniscus; Figs. 5,11,729 19). It inserts tendinously on the thickened part of the crus ventrale fossae (Figs. 7,13), 730 which slightly protrudes distally in some alcids (e.g., Cerorhinca and Fratercula; Fig. S12). 731 The mm. subcoracoscapulares complex lies deep within the caudal aspect of the 732 shoulder joint (Fig. 3). The complex has three fleshy heads: the m. subscapularis caput 733 laterale, m. subscapularis caput mediale, and m. subcoracoideus. The m. subscapularis caput 734 laterale arises from the ventral part of the lateral surface of the scapular blade between the 735 attachment sites of the retinaculum originis m. scapulotricipitis and the m. scapulohumeralis 736 caudalis (Figs. 5,11). The caput mediale has a somewhat larger origin on the cranial half of 737 the medial surface of the scapular blade and neck (Figs. 5,11). The margins of the attachment 738 sites of these two heads are only faintly delineated. The two heads are separated by the 739 aponeurosis of the m. serratus superficialis pars cranialis. In most taxa observed, the m. 740 subcoracoideus arises largely from the dorsal surface of the membrana 741 32 sternocoracoclavicularis around the processus procoracoideus (especially lig. 742 intercoracoideum). The actual attachment on the coracoid, if any, is restricted to a small area 743 of the dorsomedial aspect of the coracoidal body around the processus procoracoideus (Figs. 744 5,11). Therefore, the attachment site of the muscle is not clearly discernible on the coracoid. 745 The bellies of these heads merge into a common tendon, which then inserts into a pit on the 746 proximal aspect of the tuberculum ventrale of the humerus (Figs. 7,13). 747 The m. coracobrachialis cranialis is a short but bulky muscle on the cranial side of the 748 shoulder joint (Fig. 3). Its origin is fleshy, and lies predominantly on the dorsal surface of the 749 lig. acrocoracohumerale along its origin (Figs. 5,11). As a result, the attachment on the 750 coracoid cannot usually be traced on the bone. An exception is Spheniscus, where the muscle 751 directly arises from the ventral aspect of the processus acrocoracoideus (Fig. 19). The muscle 752 inserts on a depression (impressio coracobrachialis) on the dorsal part of the cranial aspect of 753 the proximal humerus, which lies just ventrodistal to the attachment of the m. deltoideus pars 754 minor and dorsal to the sulcus transversus (Figs. 7,13). The depression is rather small in 755 Alcidae and Spheniscus compared to the other taxa examined. 756 The m. coracobrachialis caudalis is a large muscle on the ventrocaudal aspect of the 757 shoulder joint (Fig. 3). The muscle arises fleshily from the sternal end of the coracoid, on its 758 ventral surface around the processus lateralis, and also from the adjacent lig. 759 sternocoracoideum laterale (Figs. 5,11). The medial border of the attachment site does not 760 necessarily correspond with the linea intermuscularis ventralis of the coracoid. The muscle 761 ends as a thick tendon on a distinct scar lying on the caudal aspect of the tuberculum ventrale 762 of the humerus (Figs. 7,13). In Spheniscus, the humeral insertion is displaced dorsally, and 763 lies on the crus dorsale fossae (Fig. 21). 764 The m. pectoralis is the largest and the most superficial breast muscle (Fig. 2). The 765 pars sternobrachialis of this muscle arises fleshily along the ventral part of the facies lateralis 766 33 carinae of the sternum, with the dorsal margin marked by a linea intermuscularis, as well as 767 from a large part of the lateral surface of the furcula and adjacent membrana 768 sternocoracoclavicularis (Figs. 5,11). In Gavia, the sternal origin of the pars sternobrachialis 769 extends cranially past the apex carinae by ~1 cm, resulting in a direct contact between muscle 770 fibers on both contralateral sides of the muscle (Fig. S24). The pars costobrachialis of this 771 muscle arises from the caudolateral part of the facies muscularis sterni and the associated 772 membranae incisurarum sterni; the mediocranial border of this attachment site is marked by a 773 faint ridge, which coincides with the linea intermuscularis in Alcidae (but not in the other 774 taxa examined; see below). Despite its name, the pars costobrachialis does not directly arise 775 from the rib cage in the taxa examined. The partes sternobrachialis et costobrachialis together 776 form an overall bipennate structure of the muscle; most fibers insert on the aponeurosis 777 intramuscularis, which then becomes a thick tendon of insertion. The tendon inserts on a 778 distinct scar (impressio m. pectoralis) on the ventral aspect of the distal part of the crista 779 deltopectoralis (and hence not on the entire surface of the crista deltopectoralis; Figs. 7, 13). 780 The deep side of the fascia of this muscle is also partly attached to the ventral margin of the 781 intumescentia humeri. Some of the cranialmost fibers of the pars sternobrachialis (or the so-782 called pars propatagialis) arising from the furcula do not contribute to this tendon, but instead 783 merge with the m. deltoideus pars propatagialis. 784 In Procellariidae, a distinct part of the m. pectoralis is developed, known as the pars 785 profundus (Kuroda 1960(Kuroda , 1961Meyers and Stakebake 2005). This is a pinkish, fan-shaped 786 muscle lying between the m. supracoracoideus and the main part of the m. pectoralis The m. supracoracoideus is a large, pennate muscle lying deep in the breast region 799 ( Fig. 3). It arises from the dorsocranial part of the facies lateralis carinae and the 800 craniomedial part of the facies muscularis sterni, as well as from a restricted area of the 801 membrana sternocoracoclavicularis adjacent to them (Figs. 5,11). No muscle fibers of this 802 muscle were confirmed to arise from the coracoid in the taxa examined. The attachment of 803 this muscle on the sternum is clearly bordered by distinct ridges (lineae intermusculares), 804 which form a subtriangular area in most taxa examined. The attachment site on the sternal 805 plate is usually restricted craniomedially so that it is apart from that of the m. pectoralis pars 806 costobrachialis, but it is rather extended caudolaterally in Alcidae and Spheniscus, resulting 807 in direct contact between both attachment sites. In Alcidae, the passage of this muscle 808 appears to be partly marked as a flattened scar on the medial surface of the coracoidal body. 809 The muscle turns into a thick, flattened tendon as it passes the canalis triosseus which 810 apparently acts as a pulley for this muscle. The tendon inserts on a distinct scar on the caudal 811 aspect of the tuberculum dorsale of the humerus (Figs. 7,13). In addition, the tendon is flared 812 and partly bifurcated in Charadriiformes and Gavia, inserting also on a shallow furrow on the 813 dorsal margin of the caput humeri (or the proximal aspect of the tuberculum dorsale), 814 proximally adjacent to the main insertion (see also Kovacs and Meyer 2000). 815

35
The m. latissimus dorsi complex is among the most superficial muscles of the dorsum, 816 lying superficial to the mm. rhomboideus and the scapula (Fig. 2). In most taxa examined, the 817 partes cranialis et caudalis have distinct origins, passages, and insertions. The pars cranialis 818 of this muscle is a thin, sheet-like muscle, which arises aponeurotically from the processus 819 spinosi of a few consecutive vertebrae (typically from the caudalmost cervical vertebra to the 820 cranialmost few thoracic vertebrae). The pars caudalis is somewhat bulkier, and arises as an 821 aponeurosis spanning from more caudally positioned thoracic vertebrae (typically the 822 caudalmost several thoracic vertebrae) to the area deep to the thigh musculature (e.g., mm. 823 iliotibiales); in some taxa (e.g., Catharacta and Spheniscus), the entire origin of the pars 824 caudalis lies deep to the thigh musculature. As they enter the brachium, these two parts pass 825 deep to the proximal belly of the m. scapulotriceps and cross each other with the pars 826 cranialis lying superficial (dorsal) to the pars caudalis. The pars cranialis ends fleshily on the 827 dorsocaudal aspect of the proximal humerus, along a faint ridge (linea m. latissimi dorsi) 828 extending distally from the tuberculum for the retinaculum m. scapulotricipitis (Figs. 7,13). 829 The pars caudalis turns into a tendon (or aponeurosis) which becomes closely associated with 830 the retinaculum m. scapulotricipitis. Its insertion lies just ventrodistal to that of the 831 retinaculum and ventral to the proximal margin of that of the pars cranialis (Figs. 7,13). 832 However, these attachment sites on the bone are sometimes hardly distinguishable from each 833 other. The presence of the pars metapatagialis of this muscle was only confirmed in Gavia 834 and one individual of Larus schistisagus examined in this study, in contrast to Hudson et al. 835 (1969) who stated that this part was present in all larids and alcids they examined. This part 836 might be damaged during skinning and overlooked in most birds examined here. In Gavia, 837 this part arises as an aponeurosis from the processus spinosi of the few caudalmost thoracic 838 and the cranialmost synsacral vertebrae. In Larus, the origin of this part is largely fused with 839 that of the pars caudalis. In both cases, the pars metapatagialis ends on the dermis deep to the 840 36 humeral feather tract. This muscle is highly modified in Spheniscus (see also Schreiweis 841 1982); the partes cranialis et caudalis pass through a ligamentous loop on the caudal side of 842 the shoulder joint; both parts turn into a partly fused tendon which then inserts on a distinct 843 tubercle lying distal to the crus dorsale fossae (Figs. 20,21). 844 The m. deltoideus pars propatagialis is a moderately bulky muscle on the cranialmost 845 part of the shoulder (Fig. 2). It arises fleshily from the dorsolateral margin of the furcula 846 cranial to the processus acrocoracoideus claviculae (Figs. 5,11). Just past the shoulder joint, 847 it typically turns into the common ligament of the ligg. propatagiale et limitans cubiti; in 848 Larus and Catharacta, however, the belly bifurcates before shedding the common ligament, 849 so that the two ligaments arise separately from the muscle (see above). In the taxa examined, 850 the muscle is consistently single-headed, and the delineation of multiple heads (see Vanden 851 Berge and Zweers 1993) was not confirmed. 852 The m. deltoideus pars major is a bulky muscle on the dorsal aspect of the shoulder 853 joint (Fig. 2). This muscle lies deep to the m. deltoideus pars propatagialis and typically 854 superficial to the m. deltoideus pars minor and the tendon of the m. supracoracoideus. In most 855 taxa, this muscle arises fleshily from the dorsolateral margin of the omal (dorsocaudal) end of 856 the furcula and the adjacent lig. acromioclaviculare (Figs. 5,11). In Pluvialis, Gavia,and 857 Procellariidae, the origin is slightly extended caudally to reach the acromion of the scapula 858 (Figs. S2, S24, S30). In Spheniscus, the origin seems to lie on the cranial end of the scapula 859 between the facies articularis humeralis and acromion (Fig. 19). The insertion is usually 860 fleshy (aponeurotic in Gavia), and lies on the dorsal aspect of the proximal humerus, with its 861 relative position varying substantially among taxa: in Gavia, the insertion extends proximally 862 from the middle part of the crista deltopectoralis and distally past the distal end of the crista 863 deltopectoralis (Fig. S26); in Procellariidae, the insertion lies in a depression along the entire 864 length of the crista deltopectoralis (Fig. S32); the conditions in Pluvialis, Larus and 865 Catharacta are similar, but the insertion does not extend proximally past the tip of the crista 866 deltopectoralis (Figs. 7, S4, S10); in Scolopax, where the crista deltopectoralis is relatively 867 small, the insertion largely lies on the shaft and extends as far distally as the midshaft region 868 (Fig. S8); in Alcidae, the insertion is restricted to a narrow area just craniodorsal to the 869 insertions of the retinaculum m. scapulotricipitis and the m. latissimus dorsi partes cranialis et 870 caudalis (Figs. 13, S12, S14, S18, S22). In Spheniscus, the insertion is restricted to a small 871 area just proximal to the tubercle for the insertion of the two parts of the m. latissimus dorsi 872 ( Fig. 21). 873 The m. deltoideus pars minor is a two-headed muscle of the shoulder joint with a 874 complex conformation (Fig. 3). The caput dorsale arises from the lateral aspect of the lig. 875 acromioclaviculare or the lateral surface of the furcula (in Gavia), deep (ventral) to the origin 876 of the m. deltoideus pars major. The caput ventrale lies ventral to it, arising from the 877 dorsolateral margin of the membrana sternocoracoclavicularis along the coracoid, deep to the 878 belly of the m. supracoracoideus (Fig. 11). The latter head is undeveloped in the non-alcid 879 charadriiform taxa examined (Pluvialis, Scolopax, Larus, and Catharacta). In any case, the 880 origins of these heads are not associated with any osteological correlates. A single belly is 881 formed by the two heads as the muscle passes the canalis triosseus, where the belly lies dorsal 882 to the tendon of the m. supracoracoideus. It ends largely fleshily (but partly tendinously in 883 Calonectris and Scolopax, and exclusively tendinously in Spheniscus) on a restricted area on 884 the proximal tip of the crista deltopectoralis of the humerus, with an indistinct scar (Figs. 7,885 13). 886 The m. scapulotriceps is a prominent, two-joint muscle on the dorsocaudal aspect of 887 the brachium (Fig. 2). In most taxa examined (except Spheniscus), the muscle fibers originate 888 on the retinaculum originis m. scapulotricipitis (see above); its proximal belly is anchored to 889 the dorsocaudal aspect of the humerus by a ligamentous retinaculum (retinaculum m. 890 38 scapulotricipitis), whose position is clearly marked by a tubercle on the margo caudalis of the 891 humerus, distal to the tuberculum dorsale and caudal to the crista deltopectoralis (Figs. 7,13). 892 In Spheniscus, both of these retinacula are absent (see above), and this muscle arises from 893 two separate heads on the cranial end of the scapula. One head originates from the ventral 894 margin of the scapula, and the other originates from the acromion (Fig. 19). In any case, 895 around the proximal half of the brachium, the muscle turns to a thick ligament, which then 896 passes the dorsal part of the fossa olecrani (sulcus scapulotricipitalis) of the distal humerus. 897 The tendon of this muscle ends on the proximal end of the ulna, in a depression lying just 898 caudal to the cotyla dorsalis (Figs. 9,15). 899 The m. humerotriceps is a prominent, one-joint muscle on the caudal aspect of the 900 brachium (Fig. 3). The origin is largely fleshy, and the head occupies most of the fossa 901 (pneumo-)tricipitalis on the caudal aspect of the proximal humerus (Figs. 7, 13). The head is 902 proximally incised by the crus dorsale fossae and the insertion of the m. scapulohumeralis 903 cranialis (see above) in the taxa examined except Spheniscus. In most taxa examined, the 904 portion of the head dorsal to the incision is extended so proximally that the dorsoproximal 905 margin of the attachment more or less excavates the distal margin of the caput humeri 906 (especially pronounced in some charadriiforms including Larus and Fratercula). The ventral 907 portion of the head lies deeply in the ventral part of the fossa tricipitalis. The origin may 908 extend as far distally as around the midshaft of the humerus, but its distal margin is not 909 clearly discernible on the bone. Its bellies become a thick tendon near the elbow joint, which 910 then passes the ventral part of the fossa olecrani (sulcus humerotricipitalis). The tendon of 911 this muscle is closely associated with that of the m. scapulotriceps, and these ligaments are 912 anchored to the fossa olecrani by the lig. tricipitale (see above). The tendon ends on the 913 caudoproximal aspect of the olecranon of the ulna, where the attachment is marked by a 914 prominent scar (Figs. 9,15). 915

39
The m. biceps brachii is a two-joint muscle lying deep in the cranioventral aspect of 916 the brachium (Fig. 3). This muscle appears to be absent in Spheniscus (see also Schreiweis 917 1982). Although two heads (the caput coracoideum and caput humerale) are recognized in the 918 literature, the caput humerale is apparently absent in Gavia, Procellariidae, and most alcids 919 (with the exception of Uria). The caput coracoideum arises tendinously from a distinct 920 tuberculum on the ventral aspect of the processus acrocoracoideus of the coracoid ( Spheniscus (see also Schreiweis 1982). 971 The m. flexor carpi ulnaris is a distinct, two-joint muscle on the caudoventral aspect 972 of the antebrachium (Fig. 2). The muscle arises tendinously from the distal aspect of the 973 distal extension of the epicondylus ventralis of the humerus (processus flexorius), where the 974 attachment is marked by a prominent scar (Figs. 7,13). The tendon passes the trochlea 975 humeroulnaris (see above) and then runs along the caudoventral margin of the ulna. Along 976 the antebrachium, the caudal margin of the belly (the so-called pars remigalis) is associated 977 with the lig. elasticum interremigale minor which spans the bases of the secondaries. The 978 insertion is tendinous, lying on the concavity on the caudal aspect of the ulnare (Figs. 9, 15). 979 In Gavia, the distal tendon is partly ossified. 980 The m. flexor digitorum superficialis is a thin, multi-joint muscle whose belly lies on 981 the ventral aspect of the antebrachium (Figs. 2, 3). As mentioned above, the belly of this 982 muscle arises from the deep surface of the lig. humerocarpale in the mid-distal part of the 983 antebrachium. It soon turns into a thin tendon around the wrist joint, and then passes the 984 retinaculum on the proximoventral aspect of the crus longum of the ulnare, where the passage 985 is marked by a sulcus and bony canal in some taxa (e.g.,in Larus and Catharacta;Figs. 9, 986 S10). After passing the caudal side of the processus pisiformis of the carpometacarpus, the 987 tendon of this muscle runs along the ventral margin of the os metacarpale majus, being 988 parallel and deep to that of the m. flexor digitorum profundus. As these two tendons remain 989 42 in close association along the remainder of their trajectories, it is not easy to distinguish the 990 insertions of these muscles from one another. Nevertheless, the m. flexor digitorum 991 superficialis appears to end on the ventrocranial margins of the proximal ends of either or 992 both of the proximal and distal phalanges of the major digit (Figs. 9,17). In Alcidae, the 993 insertion on the distal phalanx lies on an elongated area on the ventral margin of the bone, 994 rather than on its proximal end. In Spheniscus, this muscle is continuous with the lig. Cepphus; Figs. 15, S14, S22), or extends only slightly further distally than that impression (in 1016 Synthliboramphus, Cerorhinca, and Fratercula; Figs. S12, S18). In Spheniscus, the origin of 1017 this muscle is so closely associated with the membrana interossea antebrachii that an 1018 attachment site of the muscle separate from the membrane cannot be identified on the bone. 1019 Past the middle antebrachium, the muscle turns into a thin tendon, which passes below the 1020 aponeurosis ventralis. In the proximal manus, the tendon changes its direction on the cranial 1021 side of the processus pisiformis of the carpometacarpus, which acts as a pulley for this 1022 muscle. The tendon then goes on the cranioventral margin of the os metacarpale majus, 1023 superficial to that of the m. flexor digitorum superficialis, with which it is partly associated. 1024 The main insertion of this muscle lies on the cranioventral margin of the proximal end of the 1025 distal (second) phalanx of the major digit (Figs. 9,17). 1026 The m. extensor carpi radialis is a prominent, two-joint muscle on the craniodorsal 1027 aspect of the antebrachium ( caudale (Figs. 9,17). In Spheniscus, this muscle is present in much the same conformation, 1076 except that it completely lacks the branch leading the alular phalanx; the distal tendon of this 1077 muscle is merged with that of the m. extensor longus digiti majoris, and inserts on the 1078 proximal ends of the two phalanges of the major digit (Fig. 23). 1079 The m. extensor longus alulae is a thin muscle lying deep within the antebrachium 1080 ( Fig. 3). In most taxa examined, the muscle has two heads, on the proximal ulna and on the 1081 midshaft of the radius. The ulnar head is partly tendinous (entirely fleshy in Spheniscus), and 1082 arises from the cranial aspect of the ulna just distal to the proximal articular surfaces (Figs. 9, 1083 15); in Alcidae, this head lies in a small concavity formed by the hook-like distal extension of 1084 the cotyla dorsalis; otherwise, this origin is so vaguely marked that it is hardly discernible on 1085 the bone (perhaps except in Spheniscus). Presence of the ulnar head was not confirmed in 1086 Procellariidae and Larus. By contrast, the radial head was confirmed to be present in all taxa 1087 examined. This head arises fleshily from the caudodorsal aspect of the proximal radius, 1088 between the proximal end of the m. extensor longus digiti majoris and the distal end of the m. 1089 (interosseal) margin of the radius, but is not clearly demarcated on the bone. The two heads 1091 merge in the interosseal space, and the resultant belly crosses the dorsal side of the radius to 1092 lie within the sulcus tendinosus of the distal radius, where it turns into a thin tendon. The 1093 tendon runs alongside that of the m. extensor carpi radialis, with which it merges before 1094 inserting on the processus extensorius of the carpometacarpus. 1095 The m. extensor longus digiti majoris is a multi-joint muscle on the dorsal aspects of 1096 the antebrachium and manus (Fig. 2). The muscle seems to have two separate heads, partes 1097 proximalis et distalis. The pars proximalis arises fleshily from the caudodorsal aspect of the 1098 radius (Figs. 9, 15), typically occupying a large area of the radial shaft caudodorsal to the 1099 attachment of the m. pronator profundus (except in Larus, where the origin is restricted to the 1100 midshaft region, and in Spheniscus, where the origin is largely aponeurotic; Figs. 23, S10). 1101 The proximal margin of the origin is marked by a convergence of two lineae intermusculares. 1102 The belly turns into a thin tendon which passes the wrist joint along the dorsal rim of the 1103 trochlea carpalis of the ulna, just cranioventral to the incisura tendinosa. The much more 1104 indistinct pars distalis arises fleshily from the dorsal aspect of the proximal manus, but its 1105 origin in most taxa examined lies on the ligaments and aponeuroses spanning between the 1106 carpal bones, thus it does not typically correspond to osteological correlates. An exception is 1107 Gavia, where the pars distalis arises from the dorsal aspect of the os metacarpale majus, 1108 although the origin is only indistinctly marked (Fig. S28). In Spheniscus, the pars distalis is 1109 absent (see also Schreiweis 1982). In the proximal manus, the tendon from the pars 1110 proximalis lies along the dorsocaudal margin of the os metacarpale majus, but distally it 1111 crosses with the distal branch of the tendon of the m. extensor digitorum communis, lying 1112 superficial to the latter. In all taxa except Gavia and Spheniscus, the tendons from both parts 1113 merge to form a common tendon. When present, this common tendon then passes the cranial 1114 47 aspects of the metacarpo-phalangeal and interphalangeal joints, and ends on the cranial aspect 1115 of the proximal ends of the proximal and distal phalanges of the major digit (Figs. 9,17). In 1116 Gavia, where the major digit has three free phalanges, the tendon of the pars proximalis 1117 inserts on the second phalanx of the major digit, whereas that of the pars distalis inserts on 1118 the first (proximal) phalanx (Fig. S28). 1119 The m. supinator is a fan-shaped muscle on the dorsocranial aspect of the elbow joint 1120 Spheniscus, where that muscle does not directly attach to the ulna). 1135 As expected, the m. entepicondylo-ulnaris was absent in all taxa examined; the 1136 muscle is apparently unique to some members of Palaeognathae and Galloanseres (e.g., 1137 Vanden Berge and Zweers 1993). 1138 48 The m. ulnometacarpalis dorsalis is a fan-shaped muscle on the dorsocaudal aspect of 1139 the wrist joint (Fig. 3). The muscle seems to arise tendinously from the dorsal aspect of the 1140 distal ulna, adjacent to the protruding tubercle for the incisura tendinosa (Figs. 9,15) brachialis (Figs. 9,15). In Procellariidae, the head is slightly bifurcated, and the origin also 1151 extends caudoproximally on the ventral aspect of the distal ulna from its distal end (Fig. S32). 1152 The belly turns into a distinct tendon before entering the wrist joint from the ventral side. In 1153 the wrist joint, the tendon turns around the joint on the cranial side, along a distinct, diagonal 1154 sulcus on the cranial aspect of the radiale; here, the tendon of this muscle lies deep to those of 1155 the m. extensor carpi radialis and m. extensor longus alulae. The tendon then inserts on a 1156 distinct depression on the proximocranial part of the dorsal rim of the trochlea carpalis of the 1157 carpometacarpus (Figs. 9,17). This muscle is absent in Spheniscus (see also Schreiweis 1158Schreiweis 1982. 1159 The m. interosseus dorsalis is a small muscle on the dorsal aspect of the manus (Fig.  1160 2). The muscle arises fleshily along the dorsal margin of the interosseal sides of the ossa 1161 metacarpalia majus et minus, including the symphysis metacarpalis proximalis (Figs. 9, 17). 1162 The tendon passes a retinaculum formed on the caudal aspect of the distal end of the 1163 49 carpometacarpal shaft (at the symphysis metacarpalis distalis), and then runs along a distinct 1164 sulcus on the dorsal aspect of the proximal phalanx of the major digit. It then ends on the 1165 dorsal apex of the proximal end of the second phalanx of the major digit (Figs. 9,17); in most 1166 charadriiform taxa, the tendon also extends distally to attach on the dorsal margin of the 1167 phalangeal shaft. In Spheniscus, the tendon of this muscle runs along the caudodorsal margin 1168 of the bone, and is also attached to the proximal phalanx of the major digit (Fig. 23). 1169 The m. interosseus ventralis is another muscle on the interosseal space of the manus 1170 (Figs. 2, 3). It arises fleshily from the interosseal space of the carpometacarpus, just ventral to 1171 the origin of the previous muscle (Figs. 9,17). The origin of the m. interosseus ventralis 1172 seems to extend further distally than that of the previous muscle, extending nearly to the 1173 symphysis metacarpalis distalis. The tendon of this muscle passes another retinaculum on the 1174 dorsal aspect of the symphysis metacarpalis distalis of the carpometacarpus. The tendon runs 1175 along the caudal margin of the proximal phalanx of the major digit, and then parallel to, but 1176 separately from, the caudal margin of the second phalanx of the major digit. The tendon then 1177 attaches to a distinct eminence on the caudal aspect of the distal end of the second phalanx 1178 (Figs. 9,17), except in Spheniscus, where the eminence is absent and the tendon attaches 1179 along the entire caudal margin of the bone (Fig. 23). Before the final insertion, the tendon 1180 may be attached to the distal end of the proximal phalanx (in Larus and Cepphus; Figs. S10, 1181 S14) or the proximal end of the distal phalanx (in Catharacta, Cerorhinca, Fratercula, and 1182 Synthliboramphus; Figs. 9, S12, S18). 1183 The m. extensor brevis alulae is a small, fan-shaped muscle on the dorsal aspect of the 1184 alula (Fig. 2). The muscle arises fleshily from a depression on the dorsal aspect of the os 1185 metacarpale alulare of the carpometacarpus, particularly around the base of the processus 1186 extensorius (Figs. 9, 17). Although the margins of the origin are not clearly marked on the 1187 bone, the origin appears to be proximodistally broad in those taxa with a proximodistally 1188 50 elongated os metacarpale alulare (Gavia, Uria, Alca, and, to some extent, Synthliboramphus;1189 Figs. 17, S18, S22, S28). The muscle inserts tendinously on the craniodorsal margin of the 1190 proximal end of the alular phalanx, cranial to the insertion of the m. extensor digitorum 1191 communis (Figs. 9, 17). 1192 The m. abductor alulae is a small muscle on the cranioventral aspect of the alula (Fig.  1193 2). It arises from the ventrocaudal aspect of the tendon of the m. extensor carpi radialis near 1194 its distal end, hence its origin does not have any osteological correlates. Its belly lies along 1195 the ventral aspect of the processus extensorius of the carpometacarpus, and its insertion lies 1196 on the cranial margin of the proximal end of the alular phalanx, distal to the attachment of the 1197 lig. obliquum alulae (Figs. 9,17). In the taxa examined, with the exception of Alcidae, the 1198 insertion extends slightly distally along the cranial margin of the phalanx. 1199 The m. flexor alulae is a small, fan-shaped muscle on the ventral aspect of the alula 1200 (Fig. 3). The muscle arises fleshily from the caudal part of a depression on the ventral aspect 1201 of the os metacarpale alulare of the carpometacarpus (Figs. 9, 17). In Gavia, where the os 1202 metacarpale alulare is proximodistally elongated, the origin is restricted to the distal half of 1203 the metacarpal body (Fig. S28). The muscle ends tendinously on the dorsocaudal aspect of 1204 the proximal end of the alular phalanx (Figs. 9, 17). 1205 The m. adductor alulae is a small muscle on the ventral aspect of the alula (Figs. 2, 3). 1206 The muscle arises fleshily from the area distal to the facies articularis alularis of the 1207 carpometacarpus, along the major metacarpal shaft just distal to the ventral margin of the 1208 articular surface (Figs. 9, 17). The muscle inserts on much of the caudal aspect of the body of 1209 the alular phalanx (Figs. 9, 17). Neither the origin nor insertion can be clearly discerned on 1210 the bones. 1211 Spheniscus lacks all of the intrinsic muscles associated with the alula, including the 1212 mm. extensor brevis alulae, abductor alulae, flexor alulae, et adductor alulae (see also 1213 Schreiweis 1982). 1214 The m. abductor digiti majoris is a small muscle on the ventrocranial aspect of the 1215 manus (Fig. 3). The muscle arises fleshily from the ventrocranial aspect of the shaft of the os 1216 metacarpale majus, typically along an elongated area on the metacarpal shaft (Figs. 9, 17). In 1217 most taxa examined, the proximal margin of the origin usually reaches the area just cranial to 1218 the processus pisiformis. However, it ends on the area just distal to the processus in Scolopax 1219 ( Fig. S8), whereas it does not extend as far proximally in Cerorhinca and Fratercula (Fig.  1220 S12). In Larus and Cepphus, the origin is separated into two parts lying on the proximal and 1221 distal ends of the typical origin (Figs. S10, S14). Its insertion is tendinous, and lies typically 1222 on the cranioventral aspect of the proximal end of the proximal phalanx of the major digit 1223 (Figs. 9,17). In Scolopax, the tendon also extends to the distal end of the phalanx (Fig. S8). 1224 In Procellariidae, the insertion lies on the second phalanx, rather than the proximal phalanx, 1225 of the major digit (Fig. S32). 1226 The m. flexor digiti minoris is a small muscle on the caudal aspect of the manus (Fig.  1227 3). Its origin is fleshy, and occupies a large part of the caudal aspect of the os metacarpale 1228 minus that is not occupied by the m. ulnometacarpalis dorsalis (Figs. 9,17). Its insertion is 1229 tendinous, and lies on a prominence on the caudal margin of the phalanx of the minor digit 1230  (Figs. 32-35). It is therefore tempting to speculate that the muscle did 1262 not retain fleshy attachments in Mancalla, because fleshy attachments typically do not exhibit 1263 these sorts of osteological correlates in the taxa examined. Nevertheless, it will be impossible 1264 to definitively evaluate this idea without further analyses (e.g., histological assessment). 1265 For Pinguinus, the reconstructed musculature was subsequently compared with a 1266 dried skeletal specimen in which remnants of the elbow and forearm musculature are 1267 preserved (NHMUK 1972.1.156). The relative positions of the muscles and ligaments 1268 reconstructed from osteological correlates were consistent with those preserved in this 1269 specimen (Fig. 40), partially confirming the validity of the present reconstruction. based on parsimony tends to be more accurate for characters with low evolutionary lability 1277 (unless transition rates are biased toward derived states; Frumhoff and Reeve 1994;Schultz et 1278Schultz et al. 1996Cunningham 1999). Of course, considerable uncertainty accompanies any 1279 inferences concerning structures lacking osteological correlates (Witmer 1995)  propatagium would not be beneficial, and could even be detrimental, as water poses 1398 substantially more drag and resistance to moving wings than air, and thrust production in 1399 water is far more important than lift production (Rayner 1988). This expectation is supported 1400 by behavioral observations in extant volant auks, where the wings are seen to be kept in a 1401 partly folded position during aquatic flight, yielding smaller effective wing area than in the 1402 fully extended position (Stettenheim 1959;Spring 1971;Rayner 1995;Gaston andJones 1403 1998;Kikuchi et al. 2015). As such, both the elongated crista deltopectoralis and attachment 1404 site for the ligg. propatagiale et limitans cubiti appear to represent modifications to increase 1405 the rigidity of the leading edge of the wings, which would increase thrust production in 1406 aquatic flight (DeBlois and Motani 2019). It is evident that these modifications were acquired 1407 independently in Pinguinus and Mancallinae (Smith 2011(Smith ). 1408 In Mancalla, it is probable that function was lost in the m. size, reduced joint mobility, and reduced intrinsic musculature (Schreiweis 1982;Raikow et 1472Raikow et al. 1988Livezey 1989;Louw 1992;Bannasch 1994). At a detailed level, however, flightless 1473 auks and penguins exhibit important differences, as well as obvious similarities, in the 1474 underlying anatomical architecture of their wings. 1475 Similarities and differences are evident in the conformations of wing 1476 elevator/retractor muscles in flightless auks and penguins. In extant Spheniscidae, the m. 1477 supracoracoideus, the largest wing elevator, has a single insertion on an oblique ridge on the 1478 proximal humerus, which extends in a cranioproximal-caudodistal direction (Figs. 20, 21;1479Bannasch 1986b. By contrast, in Pinguinus and Mancalla, this muscle has a dual insertion, 1480 on the dorsal margin of the caput humeri and on the elongated tuberculum dorsale, with the 1481 latter extending along the shaft more or less in parallel with it (Figs. 26,27,32,33). These 1482 conditions may seem radically different at first glance, but are similar insofar as the insertion 1483 has extended cranially from the typical position of insertion in volant birds-the tuberculum 1484 62 dorsale on the caudodorsal aspect of the proximal humerus. The specialized conformation in 1485 the wing-propelled divers may have some mechanical consequences, especially with regard 1486 to dorsal rotation (or "supination") and retraction/adduction of the humerus, which are 1487 induced by the action of this muscle in typical flying birds (Poore et al. 1997;Tobalske and1488 Biewiner 2008). However, the exact consequences of this rearrangement will remain unclear 1489 in the absence of rigorous mechanical analyses. Notably, the conformation of this muscle in 1490 flightless auks is only slightly modified from the typical condition in Charadriiformes, in 1491 which the tendon of the muscle is partly bifurcated and its insertion extends cranially from 1492 the tuberculum dorsale (see above). The ancestral condition for Sphenisciformes was 1493 probably different, because the insertion is single in Spheniscus and in Procellariidae 1494 (representing Procellariiformes, the extant sister taxon to Sphenisciformes; e.g., Hackett et al. 1495Hackett et al. 2008Jarvis et al. 2014;Prum et al. 2015;Kimball et al. 2019). 1496 Among other wing elevator/retractor muscles, the m. scapulohumeralis cranialis, 1497 which is associated with a prominent scar on the humerus in Mancalla (above), is absent in 1498 extant Spheniscidae (Schreiweis 1982; see above). By contrast, the m. scapulohumeralis 1499 caudalis is prominent and well-developed in extant Spheniscidae, associated with the 1500 broadened scapular blade characteristic of penguins (Figs. 18, 19;Schreiweis 1982;Bannasch 1501Bannasch 1986bBannasch , 1994. This differs substantially from the rather unspecialized scapula in the flightless 1502 auks (Figs. 24, 30). Notably, the scapula is not broadened in one of the stemward-most 1503 penguins known, Muriwaimanu tuatahi, whereas it is said to be broadened in other stem-1504 group penguins including Kupoupou stilwelli, Sequiwaimanu rosieae and Kumimanu biceae 1505 (although not to the same extent as in extant Spheniscidae, and it is not well preserved in the 1506 former two taxa; Slack et al. 2006;Mayr et al. 2017Mayr et al. , 2018Blockland et al. 2019). presumably conferring only passive resilience. In addition, the processus pisiformis of the 1526 carpometacarpus-a major osseous pulley for this muscle-is lacking in extant Spheniscidae 1527 (Fig. 22). These conditions are associated with the distinctively reduced mobility of forelimb 1528 joints in Spheniscidae (Raikow et al. 1988). The processus pisiformis is present in some of 1529 the most stemward members of Sphenisciformes (e.g., Muriwaimanu tuatahi, Sequiwaimanu 1530 rosieae, and Perudyptes devriesi), whereas it has been lost/reduced in more crownward stem 1531 penguins (e.g., Icadyptes salasi, and Kairuku spp.), potentially reflecting a progressive loss of 1532 function in this muscle (Clarke et al. 2007;Ksepka et al. 2008Ksepka et al. , 2012aKsepka andClarke 1533 2010;Mayr et al. 2018). Although it is tempting to reconstruct the detailed conformation of 1534 this muscle in Mancalla based on analogy with Spheniscidae, it would be logically circular to 1535 regard any such reconstruction as evidence for convergence between these two groups. 1536 Nevertheless, based on the osteological features described above, it does seem reasonable to 1537 infer a loss or reduction of function of the m. flexor digitorum profundus in Mancalla, and 1538 therefore a reduction in the mobility of the wrist and digital joints. can acquire disparate structures specialized-or "exapted"-for the same function through 1585 multiple evolutionary pathways (i.e., non-convergent evolutionary trajectories). The resultant 1586 specialized structures, despite differing anatomically, can exhibit qualitatively similar 1587 performance for that function (Bock 1959(Bock , 1977Bock and Miller 1959), illustrating 1588 redundancy in form-function relationships. More recently, this redundancy has come to be 1589 investigated in more quantitative ways in mechanically tractable systems (e.g., Wainwright et 1590(e.g., Wainwright et al. 2005Wainwright 2007 It has previously been suggested that differing skeletal proportions in the wings of 1638 mancalline auks and penguins may have had mechanical consequences (Smith 2011; but see 1639 above for potential caveats). It has also been proposed that variability of limb skeletal 1640 proportions in birds is concentrated along the direction of clade-specific ontogeny, potentially 1641 biasing the directionality of evolutionary change (Watanabe 2018a). As such, it is possible 1642 that developmental bias precludes some wing-propelled divers from obtaining skeletal 1643 proportions that would be functionally optimal. Rigorous mechanical analyses will be 1644 necessary to fully evaluate qualitative speculations about the function of the anatomical 1645 structures discussed in the present study. Such investigations could evaluate whether the 1646 anatomical differences identified between flightless auks and penguins would have been 1647 associated with notable mechanical consequences. The avian musculoskeletal system leaves 1648 much to be explored, but synthesis of information from varied approaches, including soft 1649 tissue anatomy, the identification of osteological correlates, and mechanical analyses, will 1650 pave the way toward a more comprehensive understanding of avian morphological evolution.       layer. This illustration is partly schematic, and is not an accurate representation of muscle volume, pennation, or other architectural properties. See Table 3 for abbreviations.  Table 3 for abbreviations and Figure 2 for further information. Abbreviations: artc., articularis; impr., impressio; intermusc., intermuscularis/intermusculares; lig., ligamenti; m., musculi; proc., processus; tuberc., tuberculum.                                                views, superficial layer. The reconstruction is based on a composite of elements, thus some proportions may be inaccurate. This illustration is partly schematic, and is not an accurate representation of muscle volume, pennation, or other architectural properties. See Table 3 for abbreviations.     Major osteological landmarks mentioned in text are designated. Abbreviations: artc., articularis; impr., impressio; intermusc., intermuscularis/intermusculares; lig., ligamenti; m., musculi; proc., processus; tuberc., tuberculum.    Figure S2 for legends.             Figure S14. Osteological correlates of major wing muscles and ligaments in Cepphus carbo, wing elements. Drawn on KUGM RAJ AO13062101. See Figures S2, S4, and S13 for legends.     Figure S18. Osteological correlates of major wing muscles and ligaments in Synthliboramphus antiquus, wing elements. Drawn on KUGM RA 1311. See Figures S2, S4, and S17 for legends.