Introduction

The Late Jurassic-Early Cretaceous formations of northeastern China are well known for the extraordinary abundance and diversity, and the exceptional preservation of feathered dinosaurs that shed light on the origin and early diversification of birds (Han et al. 2014; Sullivan et al. 2014). Several small non-avian paravians (e.g. Microraptor, Pedopenna, Anchiornis, Changyuraptor, Xiaotingia, Jianianhualong) are characterized by long pennaceous feathers attached to both their tibia and metatarsus, suggesting that early bird flight might initially have involved four wings (Xu and Zhang 2005; Xu et al. 2003, 2017; Hu et al. 2009; Han et al. 2014). However, because deinonychosaurian theropods and earliest birds show a similar distribution of long pennaceous feathers along their forelimbs, hindlimbs and tail, and because different hypotheses of paravian phylogenies have recently been proposed (Hu et al. 2009; Xu et al. 2011; Godefroit et al. 2013a, b), the origin and early evolution of feather-based flight within Paraves remains controversial. Moreover, the morphology and internal structure of the feathers in the earliest paravians is poorly documented, so their real aerodynamical capacities remain conjectural. For example, the presence or absence of interlocked barbules, commonly considered as a criterion for determining whether a feather can produce useful aerodynamic lift (Zhang et al. 2006), remains unclear in the Late Jurassic basal paravians from China described so far.

Here we report a new paravian theropod, Serikornis sungei gen. et sp. nov., from the Late Jurassic Tiaojishan Formation of Linglongta (Jianchang County, Liaoning Province, China), based on a complete articulated skeleton, PMOL-AB00200, with associated integumentary structures. The plumage of this new specimen brings new information on the structure and function of the feathers in basal paravians and consequently on the early evolution of flight.

Materials and methods

This published work and the nomenclatural acts it contains have been registered in ZooBank, the proposed online registration system for the International Code of Zoological Nomenclature. The ZooBank life science identifiers can be resolved and the associated information viewed by appending the life science identifiers to the prefix http://zoobank.org/.

The life science identifiers for this publication are urn:lsid:zoobank.org:pub:65432F68-4DF9-4CC2-BE86-F366B858893A.

Results

Systematic palaeontology

Theropoda Marsh, 1881

Maniraptora Gauthier, 1986

Paraves Sereno, 1997; Avialae Gauthier, 1986

Serikornis sungei gen. etsp. nov.

Etymology

Serikos, ancient Greek for silk, because the body is almost covered with plumulaceous-like feathers; Ornis, ancient Greek for bird; named in honour of Sun Ge, for his contribution to our knowledge of Jurassic and Cretaceous ecosystems in Asia.

Holotype

PMOL-AB00200 is a single complete articulated skeleton with associated integumentary structures preserved on a slab. The counterpart is missing, but most of the skeleton is preserved on the main slab.

Locality and horizon

PMOL-AB00200 was collected in the Tiaojishan Formation (Oxfordian, Upper Jurassic; Chu et al. 2016) from Daxishan village, Linglongta (Jianchang County, Liaoning Province, China).

Diagnosis

Serikornis is characterized by the following combination of characters (autapomorphies are marked with an asterisk): four anterior maxillary teeth twice as long as the others regarding the crown height*; coracoid tuber well-developed and laterally projected from the lateral margin of the coracoid and forming a subglenoid shelf along the caudoventral margin of the bone; the distal end of the lateral process of the coracoid is thicker than the proximal part and forms a ventral rounded bump; ventrodistal process of ischium narrow, hook-like, strongly deflected caudodorsally and set at the distal end of the ischium*; smooth ventral side of coracoid devoid of small pits.

Osteological description

PMOL-AB00200 is about 49 cm long (Fig. 1). Although most of the cranial bones are shattered, the rostral part of the skull is almost well-preserved. The maxillary process of the premaxilla is short and does not exclude the maxilla from the ventral margin of the external naris, as in Archaeopteryx, Anchiornis, and most of theropods (e.g. Saurornithoides, Zanabazar, Velociraptor, Caudipteryx, Haplocheirus, Monolophosaurus, but not in Ornithomimosauria) (Barsbold 1974; Choiniere et al. 2010; Godefroit et al. 2013a; Ji et al. 1998; Makovicky et al. 2004; Norell et al. 2009; Osborn 1924; Wellnhofer 2009; Zhao and Currie 1993) (Fig. 2). The rostral plate and the maxillary process of the premaxilla are short and robust as in Eosinopteryx (Godefroit et al. 2013b). Only the ventral margin of the maxilla is well-preserved with partial shattered portions of the interfenestral bar. The left lacrimal is mostly completely preserved in Serikornis, and only the proximal part of the right lacrimal is located posteriorly to the left one. The left lacrimal has a long and slender anterior process (= rostral process), while the posterior process is shortened. It is impossible to know whether the anterior process contacts the nasal and forms the entire dorsal border of the antorbital fenestra as in all paravians where this bone is known (e.g. Aurornis, Anchiornis, Deinonychus, Sinornithosaurus, Byronosaurus) (Hu et al. 2009; Makovicky and Norell 2004; Norell and Makovicky 2004; Xu and Wu 2001). Anterior teeth are closely packed as in troodontids (Makovicky and Norell 2004; Wellnhofer 2009; Xu et al. 2011). A posteriorly widening groove extends along the labial surface of the dentary, as in many paravians (Agnolin and Novas 2013; Godefroit et al. 2013a, b; Hu et al. 2009; Makovicky and Norell 2004; Xu et al. 2011). The vertebral column is almost complete, although most details of the vertebrae are poorly preserved. Nine cervical vertebrae can be confidently identified. The second, the third and the fourth cervical vertebrae have ossified cervical ribs as long as their corresponding centra. Most of the dorsal portion of the vertebral column is covered by the left humerus, and its posterior part is shattered at the junction between two slabs. Only six dorsal centra can be identified. Details of the sacral portion is poor due to the junction of four slabs, but they appear to be fused. The tail is formed by 27 vertebrae (20 in Eosinopteryx, 30 in Aurornis, 21–23 in Archaeopteryx, 24 in Anchiornis YFGP-T5199 (pers. obs.), 22 in Jeholornis, at least 26 in Velociraptor IGM 100/986) and is 3.9 times the length of the femur (Godefroit et al. 2013a, b; Lindgren et al. 2015; Norell and Makovicky 1999; Wellnhofer 2009; Zhou and Zhang 2002). The scapula is 64% the length of the humerus (selected relative proportions are given in Table S4). This bone is thin and ribbon-shaped with a pointed distal end. It is gently convex caudoventrally whereas it is straight in Aurornis, Eosinopteryx, Xiaotingia, Microraptor, Mei, and Anchiornis (Gao et al. 2012; Godefroit et al. 2013a, b, Xu et al. 2003, 2011). The acromion process is not as strongly laterally everted as it is in Xiaotingia and oviraptorosaurs (e.g. Caudipteryx) (Osmolska et al. 2004; Xu et al. 2011), more closely resembling the moderately developed and cranially everted acromion in dromaeosaurids (Ostrom 1969; Norell and Makovicky 2004). The coracoid is subrectangular as in most basal paravians (e.g. Eosinopteryx, Aurornis, Anchiornis, Archeopteryx). The biceps tuber of the coracoid is better developed and projects further laterally in Serikornis than in other basal paravians and dromaeosaurids; it forms a subglenoid shelf along the caudoventral margin of the coracoid as in dromaeosaurids (Norell and Makovicky 2004, 1999, ). The lateral process of the coracoid is thickened distally, forming a ventral rounded bump. The boomerang-shaped furcula is robust as in most other non-avialan theropods (Xu et al. 2011). The humerus of Serikornis is slightly shorter than the femur (0.9). Its deltopectoral crest is about one quarter the length of the humerus, as in other Late Jurassic basal paravians (Table S1). The radius and ulna are straight in dorsal view, as in scansoriopterygids, Anchiornis, Eosinopteryx, and Aurornis (Godefroit et al. 2013a, b; Zhang et al. 2002). The distal end of the radius lacks the distinct flange shared by deinonychosaurians and avialans (Zheng et al. 2010). Metacarpals II and III, although partially incomplete, are not fused proximally. The dorsal margin of the ilium is gently convex as in other paravians (Xu et al. 2011) except in Aurornis, in which the dorsal margin is subhorizontal (Godefroit et al. 2013a). The obturator process of the ischium is blunt, and its ventrodistal process is strongly deflected caudorsally (Fig. 3). The femur is slightly bowed anteriorly as in Aurornis, Anchiornis, and Archaeopteryx (Godefroit et al. 2013a; Hu et al. 2009; Wellnhofer 2009). The tibia is proportionally shorter (1.41 the length of the femur) than in Anchiornis (Table S4). The metatarsals are not fused proximally, and metatarsal III is transversely compressed as in Aurornis, Eosinopteryx, Anchiornis, Microraptor, unenlagiines and all troodontids (Gao et al. 2012; Godefroit et al. 2013a, b; Hu et al. 2009; Makovicky and Norell 2004; Snively et al. 2004; Xu and Norell 2004). Metatarsal V is one third the length of metatarsal IV. Pedal digit II lacks the proximo-dorsally expanded articular trochlea of deinonychosaur-grade paravians (e.g. Deinonychus, Velociraptor) and possibly also the Eichstätt specimen of Archaeopteryx (Mayr et al. 2005; Norell and Makovicky 1999; Ostrom 1969). Pedal phalanges of digits III and IV are slender, and successive phalanges do not significantly increase in length proximodistally (see supplementary files for a complete description).

Fig. 1
figure 1

Photograph and drawing of the basal bird Serikornis sungei sp. nov. from the Middle-Late Jurassic of north-eastern China. a PMOL-AB00200 photograph. b Line drawing. Abbreviations: cev cervical vertebrae, co coracoids, cv caudal vertebrae, fu furcula, il ilium, is ischium, lf left femur, lfi left fibula, lh left humerus, lma left manus, lpes left pes, lr left radius, ls left scapula, lt left tibia, lu left ulna, pu pubis, ra radiale, rf right femur, rh right humerus, rma right manus, rpes right pes, rr right radius, rt right tibia, ru right ulna, sk skull

Full size image
Fig. 2
figure 2

Photograph and line drawing of the skull region of Serikornis. Abbreviations: de dentary, fr frontal, ib interfenestral bar, la lachrymal, ma maxillary, na nasal, pb postorbirtal bar, prm premaxillary, pt pterygoid

Full size image
Fig. 3
figure 3

Photographs and line drawings of the ischial region among several paravian theropods. a Aurornis. b Eosinopteryx. c Anchiornis. d Serikornis. Abbreviations: cv caudal vertebrae, f femur, il ilium, op obturator process, pu pubis, sac sacral vertebrae, vdp ventrodistal process. Ischia are labelled with an asterisk

Full size image

Osteological comparisons

Serikornis differs from other paravians in the following osteological features: the first four maxillary teeth that are twice as large as the others (Fig. S3 and Fig. S4). This anisodonty does not reflect taphonomic factors (i.e. a partial extrusion of first teeth from maxillary sockets) as the maxillary rim recovering the tooth is erased, allowing to observe that the roots are not displaced. X-ray analysis of the skull shows that the posterior maxillary teeth were completely erupted at the moment of the dead (Fig. S4). The biceps tuber of the coracoid is strongly developed and well detached from the lateral margin of the coracoid; the lateral process of the coracoid forms a rounded bump (Fig. S5). The ventrodistal process of the ischium is narrow, hook-like, strongly deflected caudodorsally and placed at the distal end of the bone, a unique feature in basal paravians (Fig. 3). The shape of the ventrodistal process of the ischium is not the result of a local deformation during the diagenesis because other long bones (humerus, ulna, radius, femur and tibia) remain straight along their entire length.

Although they share relatively close bone proportions (Table S4), several characters clearly distinguish Serikornis from Anchiornis, also from the Tiaojishan Formation. The ventral side of the coracoid in PMOL-AB00200 is not sculptured by the numerous small pits, regarded as autapomorphic for Anchiornis (Fig. 4) (Lindgren et al. 2015; Makovicky and Norell 2004; Xu et al. 2009). As this character is present both in the smaller holotype (femur length 43 mm; Xu et al. 2009) and in a larger referred specimen (YFGP-T5199, femur length 70 mm, as in the holotype of Serikornis; Lindgren et al. 2015) of Anchiornis, the ornamentation of the coracoid is clearly not ontogenetic. The ischium of Serikornis differs from that of Anchiornis by its blunt obturator process located more distally than the midpoint of the ischial shaft (Figs. 3c, d and S6). In Anchiornis, the posterodorsal margin of the ischium is gently concave in lateral view (Hu et al. 2009) (pers. obs. on YFGP-T5199). This is the plesiomorphic paravian condition, also shared by Eosinopteryx (Godefroit et al. 2013b). The ischium of Serikornis is also unique among basal paravians because of its hook-like process on the apical end of its dorsal margin; this process is clearly visible on the left ischium, and as an impression on the slab above the damaged end of the right ischium that closely resembles the process on the left ischium (Fig. 3d). In both ischia, the process is more expanded dorsoventrally than anteroposteriorly, resulting in a narrow ‘hooked’ apex of the bone. Serikornis also differs from Anchiornis in having a tibiotarsus/femur ratio smaller than 1.5 (1.41 in YFGP-T5199, > 1.55 in all known Anchiornis specimens regardless of body size (Hu et al. 2009; Xu et al. 2009; pers. obs. on YFGP-T5199, Table S3)). This ratio is shared by Anchiornis specimens smaller than Serikornis (pers. obs., on YFGP-T5199) and individuals subequal in size to the latter (femur length 66 mm; ratio = 1.60; LPM-B00169) (Hu et al. 2009). Therefore, different proportions of the tibia in Serikornis and Anchiornis likely do not reflect ontogenetic variations.

Fig. 4
figure 4

Comparison of the scapular girdle in Serikornis sungei and Anchiornis huxleyi (YFGP-T5199). a Serikornis sungei. b Anchiornis huxleyi. Numerous pits are present on the anterior surface of the coracoid of Anchiornis that are completely absent in Serikornis. Abbreviations. co coracoids, ct coracoid tuber, fu furcula, lh left humerus, lpr lateral process of the coracoid, ls left scapula, rs right scapula

Full size image

Serikornis differs from Aurornis in the morphology of its ischium: the postacetabular process tapers distally and the dorsal margin is gently convex, whereas it is not markedly deflected ventrally with a horizontal dorsal margin in Aurornis. The ischium also lacks the ventrally expanded, hook-like ventral process on the distal end that delimits a prominent distal obturator notch in Aurornis (Godefroit et al. 2013a).

Serikornis differs from Eosinopteryx in the longer rostral plate and larger maxillary process of the premaxilla (although this latter difference may be biased by the different ontogenetic stages of the type of individuals). It also differs by the presence of an anterior process of the lacrimal (the latter is vestigial in Eosinopteryx). This feature is likely not an ontogenetic feature as Serikornis is a subadult specimen whereas Eosinopteryx is thought to be an earlier juvenile (Godefroit et al. 2013b).

Serikornis differs from Pedopenna in its relatively longer metatarsal I, its metatarsal III longer than metatarsals II and IV, and in the more robust first phalanx of pedal digit I (Xu and Zhang 2005).

Serikornis differs from Xiaotingia in the posterior process of its maxilla that does not exceed in depth the dentary at mid-length, in having a metacarpal III as long as and not more robust than metacarpal II, in the interclavicular angle of its furcula wider than 75° with a slender omal end, and in lacking a strongly laterally everted acromial process on its scapula that overhangs a groove along the lateral surface of the scapular blade (Xu et al. 2011).

Limb element ratios used to differentiate Serikornis may be affected by allometric effects as in other theropods (Houck et al. 1990; Gatesy 1991; Christiansens 1999; Currie 2003). In order to demonstrate that the difference in proportions is due to taxonomy rather than ontogeny, a plot of the limb ratios presented in Table S4 versus the femur length (taken as a body size proxy) is presented in Fig. 5. Trend lines indicate that there is no relationship in regard of each limb proportion ratio. Therefore, the differences in limb proportion ratios are due to taxomony rather than to ontogeny and they can be used in order to discriminate Serikornis from other paravian species.

Fig. 5
figure 5

Plot of limb proportion ratios versus the length of the femur as a body size proxy in seven paravians. Trendlines and R 2 indicate an absence of relationship between species in regard of each limb proportion ratio

Full size image

Pneumaticity

Without the presence of recognizable foramina in Serikornis due to the crushing of bones during diagenesis, it is impossible to determine whether the bones are pneumatized. Thanks to a X-ray technique called laminography (see supplementary files), the skull and the four anteriormost cervical vertebrae were digitally sliced with high resolution (Fig. 6). At least the anteriormost cervical vertebrae are pneumatized and form numerous small internal chambers (camellate vertebrae). This pattern suggests that a cervical air sac system invades those vertebrae (Sereno et al. 2008) as it is also the case in many theropods (e.g. Aerosteon, Archaeopteryx, Rahonavis, Deinonychus, Velociraptor, Anchiornis) (Britt et al. 1998; Forster 1998; Norell and Makovicky 1999; Ostrom 1969; Sereno et al. 2008; Xu et al. 2009).

Fig. 6
figure 6

Pneumaticity in Serikornis sungei under laminography. At least the most proximal portion of the cervical column is highly pneumatized as shown by numerous small chambers revealed under laminography. Abbreviation: cev cervical vertebra

Full size image

Plumage description

Feather impressions are present around the entire skeleton, extending onto the pedal phalanges (except the unguals), as also described in Anchiornis (Hu et al. 2009) and Xiaotingia (Xu et al. 2011) (Fig. 7). It is not possible to observe whether the long and thin filaments that cover the top of the skull remain isolated or are bundled proximally. Feathers similar to those previously identified in the dromaeosaurid theropod Sinornithosaurus (Ji et al. 2001) cover the neck of Serikornis and consist of bundles of filaments that are joined together proximally and remain nearly parallel as they extend distally (down-like feathers) (Fig. 7a). The back of Serikornis is entirely covered by contour-like feathers with well-defined rachis and transversely inserted barbs (Fig. 7b). As in Anchiornis, the forelimb wings appear to be formed by multiple layers of relatively short, slender, symmetrical and poorly differentiated feathers inserted on a large propatagium; unlike in Archaeopteryx, Microraptor, Confuciusornis and modern birds, remiges and coverts cannot be distinguished (Fig. 7c) (Brown et al. 1994; Chiappe et al. 1999; Li et al. 2012; Longrich et al. 2012). The number of pennaceous feathers attached to the forelimb cannot be accurately estimated. As in Anchiornis (Hu et al. 2009) and Eosinopteryx (Godefroit et al. 2013b), the longest forelimb feathers are attached near the distal end of the forearm and the proximal end of the manus, and are about 150% the humeral length. A large number of undifferentiated forelimb feathers are arranged in several rows in Serikornis, suggesting the presence of a large propatagium, a fleshy structure made of skin filling the angle between the humerus and the forearm (Baumel 1993; Brown et al. 1994; Feduccia and Czerkas 2015), in life. In modern birds, the propatagium is mainly responsible for the aerodynamical profile of the wing (Brown et al. 1994; Lucas and Stettenheim 1972), but this structure is also necessary for the insertion of numerous covert feathers responsible for the cambered dorsal surface of the wing (Brown et al. 1994). The development of a large propatagium in basal paravians is therefore supported by the presence of numerous layers of feathers along the forearm in Serikornis but also in Anchiornis (Longrich et al. 2012; Wang et al. 2017), although this character could be inherited from a common ancestor with oviraptorosaurs as this character is also present in Caudipteryx (Feduccia and Czerkas 2015). In Serikornis, the vanes of the remiges obviously lack barbules, as observed in both optical and electronical microscopy (Fig. S8 and Fig. S9), and confirmed by the absence of basal expansions along the barbs that mark the insertion of barbules in modern feathers (Lucas and Stettenheim 1972); however, those basal expansions have not been observed either in early bird specimens assumed to have barbules (see supplementary files). Of course, it cannot be excluded that this apparent absence of barbules is a taphonomic bias. However, thin integumentary structures, with a mean width of 43 ± 0.002 μm, similar to the range of modern barbule widths (e.g. 23 μm in the shelduck up to 80 μm in the single comb chicken; Lucas and Stettenheim 1972; D’Alba et al. 2014), are preserved in the neck region indicating that the absence of barbules does not reflect preservation conditions. If barbules were really missing, the pennaceous feathers of Serikornis did not form an efficient aerial surface for supporting the air pressure during a wing-propelled flight (Lucas and Stettenheim 1972). Ji et al. (1998) argued that the presence of a well-formed vane in fossilized feathers, as observed in Serikornis, is an indirect evidence for the presence of barbules. However, the distal part of the covert feathers of the cassowary (Casuarius casuarius) does not form a loose tangle of barbs although they are devoid of barbules (Fig. S10). Moreover, the proximodistal distances between the proximal barbs in this bird are sometimes insufficient to allow interlocking barbs. It is therefore conceivable that complete interlocked vanes evolved later in paravian evolution, when the main role for feather changed from a social/thermoregulatory display to a flight function.

Fig. 7
figure 7

Integumentary structures of Serikornis sungei sp. nov. a ‘Down’ feathers from the neck region. b ‘Contour’ feathers from the back region. c Symmetric feathers of the left wing where the remiges and the contour feathers cannot be distinguished. d Tibial feathers on the posterior face of the right femur. e Metatarsal feathers recovered after a second preparation. f Distal part of the tail covered by numerous ‘down’ feathers. g Upper part of the tail where pennaceous symmetric feathers are inserted. Scale bars: a–d, f, g = 10 mm and e = 5 mm

Full size image

Densely packed integumentary filaments that are joined together proximally are present along the posterior part of the hindlimbs—from the femur up to the penultimate pedal phalanges—contrasting with Eosinopteryx in which the lower leg is devoid of any integumentary structures (Godefroit et al. 2013b), although it cannot be excluded that this character reflects taphonomical conditions. Besides the down-like feathers, meticulous preparation of PMOL-AB00200 reveals the presence of tibial and of short metatarsal remiges (Fig. 7d and Fig. S11), as in Anchiornis, Archaeopteryx and Sapeornis (Foth et al. 2014; Xu et al. 2003; Xu and Zhang 2005; Zheng et al. 2013b). Serikornis can be viewed as a tetrapterygian paravian (= four-wing biplan made of elongated feathers), although its forewing is composed of undifferentiated contour and flight feathers, and its hindwing includes both pennaceous and plumulaceous-like feathers (both wings may lack barbules). As in Anchiornis, the presence of both plumulaceous-like feathers and pennaceous feathers on the hindwing of PMOL-AB00200 represents a transitional stage between the fully plumulaceous leg of basal coelurosaurians (e.g. Sinocalliopteryx and Yutyrannus) and a fully pennaceous hindlimb recovered in more derived Avialae (e.g. Archaeopteryx and Sapeornis) (Godefroit et al. 2013a; Hu et al. 2009; Zheng et al. 2013b).

The tail of Serikornis is proximally covered by numerous down-like feathers (Fig. 7e), while short pennaceous symmetric feathers with a slender rachis are inserted along the distal end of the tail (Fig. 7f). The distal tail feathers of Serikornis cannot be considered as true rectrices (that is, large-sized asymmetrical feathers of the tail) but more closely resemble tectrices (upper tail coverts) in Avialae (O’Connor et al. 2013). As in the hindlimb, the tail feathers of Serikornis therefore represent a transitional condition between the fully plumulaceous tail filaments of more basal coelurosaurs and the longer rectrices inserted all along the tail of Anchiornis (see YFGP-T5199 specimen in Lindgren et al. 2015) and basal Pennaraptorans (Foth et al. 2014; Hu et al. 2009).

Phylogenetic analysis

The data matrix consists of 1732 informative characters, including 9 characters describing the plumage, and has been scored for 130 operational taxonomic units (OTUs) (see supplementary files). In order to recover the position of Serikornis, the data matrix has been analysed with the TNT software (Goloboff et al. 2008). One hundred ‘New Technology search’ runs with default settings have been computed. Then the shortest tree islands found were explored by the tree-bisection-reconnection algorithm to only hold the shortest trees. Our analysis recovered 96 shortest trees with each a length of 5743 steps. The consensus tree focused on the paraves clade (Fig. 8) shows that the scansoriopterygid lineage is placed at the base of the clade. Here, Xiaotingia is recovered as a basal member of Scansoriopterygidae but its position is highly unstable (Foth et al. 2014; Godefroit et al. 2013a, b; Xu et al. 2011) and further investigations are required to fix its position. Several taxa (Pedopenna, Aurornis, Eosinopteryx and Anchiornis) previously regarded as basal members of Avialae are now placed in a more basal position outside Eumaniraptora. As in numerous studies, Troodontidae and Dromaeosauridae are grouped within the Deinonychosauria, which is the sister group of Avialae.

Fig. 8
figure 8

Strict consensus tree resulting from the phylogenetic analysis (length = 5743, consistency index = 0.2614, retention index = 0.7386). Major evolutions in the history of birds are mapped on the paraves clade. The complete tree is presented as supplemental fig. S11

Full size image

Our phylogenetic analysis places Serikornis among basal paravians, outside the Avialae-Deinonychosaurian node (Eumaniraptora) (Fig. 8 and Fig. S11), as the sister taxon of Eosinopteryx, and closely related to Aurornis and Pedopenna (Foth et al. 2014; Godefroit et al. 2013a; Hu et al. 2009; Lefèvre et al. 2014). Our analysis also recovers Anchiornis as the sister taxon of Eumaniraptora.

Discussion

Our results are consistent with the presence of four wings as the primitive condition for Eumaniraptora (Fig. 8) and inherited by basal birds or, in other words, that the flapping flight of modern birds was preceded by a four-winged gliding stage (Godefroit et al. 2013a, b; Longrich et al. 2012; Xu et al. 2003, 2011; Xu and Zhang 2005; Zheng et al. 2013b). Fully developed hindlimb wings, implying the presence of elongated remiges along both the tibia and the metatarsus, are present in Microraptor (Xu et al. 2003), Pedopenna (Xu and Zhang 2005), Anchiornis (Xu and Zhang 2005), Changyuraptor (Han et al. 2014) and Sapeornis (Zheng et al. 2013b). Although the hindlimbs of Serikornis are covered both by bundles of filaments joined proximally and by fully developed pennaceous feathers, this pattern remains consistent with the tetrapterygian condition of basal birds. Eosinopteryx seems to be devoid of hindlimb wings: pennaceous feathers are only present along the posterior part of the thigh and crus, but this absence can be the result of a taphonomic bias. This apparent reduction of the hindlimb plumage may be regarded as a secondary loss in Eosinopteryx (Godefroit et al. 2013b) or as an ontogenetically controlled feature.

The symmetrical vanes on the hindwing feathers of Serikornis, Anchiornis and Pedopenna (Xu and Zhang 2005) seem to be less efficient from an aerodynamic perspective than those of more derived paravians, so these taxa may have used their hindwing feathers for other functions, such as visual display or mate recognition. The high plasticity in the development of metatarsal plumage in paravians can still be observed in modern birds, with the recurrent development of feathered feet in birds of prey (e.g. Aquila chrysaetos, Asio flammeus, Bubo scandiacus) or in chicken breeds such as Silkies (Barrows 1981; Bartels 2003). Although their small size suggests that these animals were probably not top predators, they would have needed fast movements to escape predation. In this way, the development of hindwing feathers remains disadvantageous and should be regarded as a sign of sexual selection (Chiappe et al. 1999; O’Connor and Chang 2015).

Although elongated rectrices are present along the distal half of the tail in Oviraptorosauria (Ji et al. 1998; Xu et al. 2010) and Microraptor (Li et al. 2012), they seem to be absent in Serikornis, Aurornis and Eosinopteryx, which may suggest that the tail of those basal paravians had no aerodynamic function in increasing the total lift of the animal while gliding as in Archaeopteryx (Longrich et al. 2012), nor any display function as well. The long bony tail was completely covered by elongated pennaceous rectrices in both Anchiornis (Hu et al. 2009) and Archaeopteryx (Foth et al. 2014), whereas Jeholornis had both proximal and distal tail fans (O’Connor et al. 2013). The tail plumage is highly variable in pygostylian birds: a pair of elongate rectrices in Confuciusornis male specimens (Chinsamy et al. 2013) and, in some Enantiornithes (Wang et al. 2014), a graded fan of pennaceous feathers in Sapeornis, a forked tail of pennaceous feathers in Schizooura and a fan-shaped tail of pennaceous feathers in Hongshanornis and in most modern birds (Lucas and Stettenheim 1972; Wang et al. 2014). Rectrices are secondarily absent in Confuciusornis females (Chinsamy et al. 2013) and some enantiornithine female birds (Foth et al. 2014; Zheng et al. 2013a).

Although they extensively covered both arms and hands, the forelimb feathers of the Middle-Late Jurassic basal paravians Anchiornis, Eosinopteryx and Serikornis remained unspecialized and undifferentiated into elongated remiges and shortened coverts, unlike in Archaeopteryx, Microraptor and modern birds (Foth et al. 2014; Longrich et al. 2012): all were rather short, slender, symmetrical and, at least in Serikornis and Eosinopteryx, devoid of barbules, contrasting with the more elongated and asymmetrical wing feathers with well-developed barbules in Archaeopteryx (Carney et al. 2012; Foth et al. 2014) and modern birds (Lucas and Stettenheim 1972). Based on these results, the forelimb feathers in Late Jurassic basalmost paravians were obviously not adapted for active flight (although glide flight cannot be excluded) and were therefore more likely related to other biological phenomena, including visual display and sexual selection (Foth et al. 2014; Ji et al. 1998; Li et al. 2010; O’Connor et al. 2013; Prum and Brush 2012; Xu and Guo 2009).

The supposed limited flight capabilities of Aurornis, Eosinopteryx and Serikornis, as evidenced by the study of their preserved plumage, is also reflected in their osteology. Their forelimbs are proportionally shorter and more gracile than in Anchiornis and Archaeopteryx (Table S1-S2), resulting in a reduced wing surface. All these specimens also have a relatively straight ulna and radius (thus limiting the pronation and supination movements necessary for producing a wing beat), lack a bony sternum for attachment of powerful pectoral muscles and have proximodistally decreasing pedal phalanges, together with small pedal unguals that are poorly recurved (Pike and Maitland 2004). Moreover, the relatively high ratio of tibiotarsus length to femur length can be regarded as a good evidence of cursoriality (Boles 1997). All these characters suggest that basal paravians were primarily ground-dwelling animals with good cursorial abilities (Foth et al. 2014; Hu et al. 2009). However, the manual digits of Serikornis are long and slender with strongly curved unguals I and III. This supposed that they could have been effective for climbing trees as in Archaeopteryx (Feduccia 1993; Manning et al. 2009; Wellnhofer 2009). In this way, the hindlimbs can be regarded as less specialized than the forelimbs for grasping.

The recent discovery of a patagium in scansoriopterygids (a lineage found among basalmost paravians in our phylogenetic analysis, and lacking evidence of remiges) (Xu et al. 2015) suggests that the earliest adaptation to an arboreal/gliding lifestyle among paravians did not involve exaptation of the plumage as an aerodynamic surface. This is particularly true as the insertion of several rows of forelimb feathers requires a large propatagium. The absence of true flying adaptation in the ‘tetrapterygian’ Serikornis and the gliding membrane of scansoriopterygids both challenge an aerodynamic function as earliest driver of plumage elaboration in basalmost Paraves.

Birds and, by extension, some other archosaurs are characterized by a pneumatic postcranial skeleton with invasion of bones by the pulmonary air-sac system (Benson et al. 2012; Britt et al. 1998; O’Connor and Claessens 2005). This system allows a flow-through ventilation and exceptionally efficient gas exchanges (Duncker 1971), and has two evident additional functions: weight reduction in large-bodied or flying taxa and density reduction by energetic savings during foraging and locomotion (Benson et al. 2012; Bramwell and Whitfield 1974; Britt 1993; Cope 1877; Currey and Alexander 1985). The latter function is widely accepted as the main reason for skeletal pneumatization because body size has no significant influence on the proportion of pneumatized skeletal compartments (O’Connor 2004).

Pneumatic foramina (that is, the opening that allows an air sac to enter bone) are proportionally much larger than the primitive nutriment foramina in non-pneumatic vertebrae (including apneumatic bird vertebrae) (Britt et al. 1998). Pneumatic foramina are present in most tetanuran theropods (Benson et al. 2012; Britt et al. 1998), dromaeosaurs (Makovicky et al. 2005; Ostrom 1969), oviraptorosaurs (Osmolska et al. 2004), birds (Apostolaki et al. 2015; Baumel 1993; O’Connor 2004), sauropods (Cope 1877; Marsh 1877; Upchurch et al. 2004; Wedel 2003) and pterosaurs (Bonde and Christiansen 2003; Butler et al. 2009; Claessens et al. 2009; Eaton 1910; Seeley 1870). The character distribution of pneumaticity shows that although axial pneumaticity may lighten the skeleton, its evolution cannot be considered to be an adaptation for flight (Britt et al. 1998). The pneumaticity of cervical and anterior dorsal vertebrae occurred early in theropod evolution (Benson et al. 2012), and the presence of pneumatic foramina in vertebrae of non-avialan and avialan theropods indicates that some components of the avian air-sac lung system was already, to some degree, in place (Britt et al. 1998).

Britt (1993) proposed several osteological correlates of vertebral pneumaticity, based on osteological study of extant ratites (Struthio camelus and Dromaius novaehollandiae) (e.g. large external foramina, external fossae with a crenulate surface texture, thin outer bone walls) (Benson et al. 2012). However, O’Connor (2006) noted that several of these features are present in crocodilians, which lack postcranial pneumaticity. Thus, the presence of internal chambers (called camerate or camellate based on the number and size of internal chambers) opening externally via large (and thus not simply vascular) foramina is the only unambiguous evidence of skeletal pneumaticity (Britt et al. 1998; O’Connor 2006; Wedel 2007). The presence of a high pneumaticity of the anteriormost cervical vertebrae and the limited flight capacities of Serikornis suggest that high pneumatization in small maniraptorans reflects the demands of an increasingly high-performance metabolic regime rather than a prerequisite for flight (Benson et al. 2012; Britt et al. 1998; Cubo and Casinos 2000; Currey and Alexander 1985; Fajardo et al. 2007; O’Connor 2004). Further analyses are required to explore the pneumaticity of the whole specimen. The laminography technique is promising as it allows to investigate the proportion of pneumaticity without external traces of foramina or pneumaticity.