Previously, several authors have noted the presence of body feathers in Archaeopteryx (London specimen [90]; Berlin specimen [4], [26], [30], [33], [49] and [102]; Archaeopteryx in general [1], [119] and [143]) – although [1] and [143] undoubtedly referred to the Berlin specimen) –; Maxberg specimen [51] and [107]). What appears in early drawings to represent faint impressions of feathers along the lower part of the back has usually been only casually discussed, for instance by [26] and [30]: “the tail coverts form a continuous series with the body feathers on the back” [30 (pp. 37–38)]. It is unclear if the remark by Ostrom [86 (p. 118)] referred to the presumed body plumage or the flight feathers. The inference of body feathers in the London specimen is incorrect ([30] and pers. obs.), and the Maxberg specimen is missing. The plumages of the other specimens are much more poorly preserved, leaving only the Berlin specimen relevant for study.

In the original description of the Berlin Archaeopteryx, Dames [26 (pp. 39–42)] provided the most detailed account of its possible body feathers. He described the feathers around the base of the neck and of the tibia, the latter of which he thought were somewhat ellipsoid in outline and more simple and soft than the rectrices. Heller [51] discussed the feathers on the legs in comparison with corresponding ones in the Maxberg specimen, which are symmetrical, but more oval than those of the Berlin specimen. Feduccia [36 (p. 113)] stated that the latter were “typical contour feathers”. Gauthier and de Queiroz [44 (p. 13)] even thought that the legs feathers in the Berlin Archaeopteryx had open vanes, i.e. lacking hamuli, although they offered no evidence for this claim. It does, however, appear to be correct, as noted below. Elzanowski [32] also accepted the feathers on the lower leg as “contour feathers”, but did not describe them.

There is also indirect information in the innumerable illustrations made of Archaeopteryx over the years, since they presumable reflect the current opinion of the distribution and morphology of its plumage. Many early authors illustrated Archaeopteryx with feathered epipodials (e.g., Fig. 3, [44 (frontispiece and Figs. 2, 50, 64: pp. 92–93), 91 (Figs. 3, [102] and [105]]. This is similar to some more modern reconstructions (e.g., Hayward in [30], Freund in [94], Sibbick in [82], Sovak in [23 (pp. 99, 111)], [39], Nichols’ in [132], Desselberger in [32], Rey in [21 (front cover)], and in [45]. However, the majority of reconstructions depict Archaeopteryx with short feathers on the legs (e.g., Charles Knight in his classical ‘murals’, see e.g. [96], Sibbick in [75], [83] and [132], Bakker [2] and [3], Thulborn and Hamley [123], Paul in [95] and [97]. Feduccia [36] illustrated Archaeopteryx with a sort of narrow “feather trousers”(but also with a ‘tuft’ on its head, for which there is no evidence) A detailed and critical review of earlier attempts to reconstruct Archaeopteryx is given by Stephan [119], but ends (Fig. 107) with an illustration lacking any indication of longer feathers on the lower legs, and throughout his otherwise very detailed book, these feathers are not described, only mentioned once (p. 68).

Short leg feathers are frequently attributed to the hypothetical reconstructions of a ‘proavian’ stage, (e.g., [87] and [122]), whereas Heilmann [48] and [49] endowed his ‘Proavis’ with a sort of short “feather trousers”, a single line of slightly longer feathers posterior to the femora, and more narrow feathers on both sides of the tibiae, the former equivalent to the so-called “pelvic or femoral wing” (pp. 195–196) inspired by the Tetrapteryx idea of Beebe [4]. Heilmann’s feather frills along the neck and body appear unsubstantiated both in fossils and in embryology, but seem inspired by the lateral ‘skin-folds’ of the gliding gecko (Fig. 141). His “Proavis” resemble that of [115].

Pycraft’s [103] Proavis had short arms, emerging remiges, and scaly legs, and he depicted Archaeopteryx with longer “feather trousers” on the lower legs. Steiner’s [115] lizard-like Proavis (nearly identical picture in Feduccia [36 (p. 92)], with small leg-scales is similar to Heilmann’s, but the distally narrow, tapering tail-fan, also seen in Pycraft [103], is contrary to current fossil evidence of fan-like rectrices in avialean outgroups, as noted below. The rectrices appear to evolve from the broader distal to the proximally narrower part of the tail, as predicted in Tarsitano’s [122] parachuting to gliding proavian stages. All of these glider models are, however, erroneous in having rather long tertiaries, which is unsubstantiated both from fossils [40] and [118], and nearly all recent and embryonic birds. This also applies to Bakker’s [2], Thulborn & Hamley’s [123] and Paul’s [97 (Fig. 9.1)] Archaeopteryx models. A more convincing ‘Proavis’ was reconstructed by Paul [97] as a stem-Paraves (sensu [110]), Based on current fossil evidence, it would appear that a stem-avialean ‘Proavis’ would have had much longer feathers on the hind legs, possibly even ‘wing-like’ (cf. [123] and [142]).

Although the Berlin specimen, some years after its description, was named Archaeopteryx siemensis [27], and [98] referred it to Archaeornis, we here, despite the slight (10–15 %) size differences, consider it conspecific with the London specimen, Archaeopteryx lithographica von Meyer 1861 ([74]; assuming that this is the proper name for the London skeleton – Opinion 1070 [84]; contra [32], the species under consideration here). The history behind von Meyer’s name can be found in [30] and [134].

From a taxonomic viewpoint it is worth noting that the taxon Archaeopteryx, as normally perceived, is probably not a monophyletic group, but rather encompasses several species, perhaps even genera, whose exact interrelationships are presently unknown [9]. Apart from Archaeopteryx lithographica, the most significant taxa suggested for other specimens are Jurapteryx recurva (Eichstätt specimen, [59]), Archaeopteryx bavarica (the 7th specimen, [130]), and the hitherto largest known specimen, the 6th or Solnhofen specimen [129] has recently been selected as type specimen of a new genus and species, Wellnhoferia grandis [31] and [32]. Both the interrelationships of these taxa, and the relationship to later avialeans are unclear, but it is clear that the taxon Archaeopteryx sensu lato (for all the specimens from the Lithographic limestone) is a so-called ‘mixotaxon’, not necessarily monophyletic [9] (despite [32]).

Thus, it follows that these specimens need not have been uniform in ecology and soft tissue anatomy. This has implications for inferences of plumage, because what is present in the Berlin specimen needs not have been similar in all species.

Witmer [137], in his phylogenetic bracket methodology, has argued in detail how to infer the presence of soft tissue structures in fossils in a cladistic framework by reference to the living sistergroup and their combined living outgroup. The broader, general conditions and arguments for estimating such unknown features (or as the case were, the palaeoecology) on a well-known phylogenetic background were earlier outlined by Bonde [8]. This requires that soft tissue inferences pass both the test of morphological similarity and phylogenetic congruence.

For every species of Archaeopteryx** (see [9] and Elzanowski’s ‘Archaeopterygidae’ [32]) the recent sistergroup is Aves and the outgroup Crocodylia, making it impossible to infer whether the fossils had feathers or not, as these do not usually leave distinct osteological signs, with the exception of quill nodes or a pygostyle. However, in the present case fossil preservation allows prediction with certainty that at least some and probably all specimens had feathers in the form of remiges and rectrices. This, however, implies neither the unquestionable presence of body plumage (see, e.g., Bakker [2 (p. 69)]), nor the resemblance of these, if present, to those of modern birds. Only fossil discoveries in an outgroup position to birds plus Archaeopteryx (= Avialae) with preserved body feathers would make such inferences possible.

Phylogenetic analyses have demonstrated with overwhelming certainty (Fig. 4) that Avialae is the sistergroup of advanced theropod dinosaurs, most often dromaeosaurs (the two groups combined in Paraves, sensu Sereno [110]). Since the mid 1990s, theropods have emerged from Early Cretaceous beds of Liaoning, northeastern China, with primitive, superficially hair-like feathers preserved on their bodies [15], [62], [138] and [139], and some with down-like structures ([93] and NB’s pers. obs.) and even with large remiges and rectrices [16], [61], [62], [78], [81], [150] and [151], despite being non-Avialan theropods [21], [24], [54], [55], [61], [79], [80], [110] and [120].

The hair-like body feathers are clearly different from mammalian hair [15], [25] and [62] and probably correspond to early evolutionary stages of ‘proto-feathers’ [10] and [141] (see also Prum [100]). Thus, they would be equivalent to ‘cryptoptiles’ [10], a term used by neornithologists (e.g., [22] and [124]) for a hypothetical early stage of feather evolution. Curiously, feathered theropod dinosaurs were not only foretold, but also outright advocated by Lowe [65], who also considered Archaeopteryx simply a feathered, arboreal dinosaur unrelated to birds!

A very small dromaeosaurid has structures preserved behind the hind legs that resemble elongate contour feathers [141], and more recently described dromaeosaurids have very large, distinctly pennaceous feathers along the forelimbs, hind limbs and tail [45, 101, 81 (Cryptovolans), 142 (Microraptor)]. Feather morphology, especially the asymmetrical remiges, in these forms suggest gliding or even flying abilities, although more analyses are required to firmly establish this [91]. Microraptor [142] also appears to possess not only Archaeopteryx-like wings, but apparently distinct ‘hind limb wings’, providing some credibility to Beebe’s [4] Tetrapteryx hypothesis [10] and [142].

To make a more definite statement about possible hind limb wings in stem-avialeans, i.e. the direct forerunners of birds, we need information from well-preserved feathered hind limbs of fossils in sistergroup position to Paraves, as it cannot be ruled out that ‘four wings’ might be an apomorphy of primitive dromaeosaurs [101]. To date fossils in the desired position are oviraptorosaurs, and among these Caudipteryx seem to have short feathers on the lower legs and nothing on the feet ([61], [149] and [150] and NB’s pers. obs.]. The inference for the stem-bird would be absence of ‘hind limb wings’, unless Archaeopteryx does, in fact, show something else (below). Studying other early birds with plumage (partly) preserved (like Confuciusornis [20]) is another possibility to establish a ‘phylogenetic bracket’ of extinct animals around the origin of avialans. But the fact that these early birds from China [57] seem to exhibit no feathers on the feet and only rather short ones on the lower legs (e.g., [20] and [146]), i.e. ‘modern’ features like most living Aves, does tend to support the deduction above: no Tetrapteryx stage in the direct lineage towards modern birds (contra [142], see also [91]).

Many confuciusornithid birds from these Chinese deposits have the plumage partly preserved, often showing very long and narrow, asymmetrical primaries, but the rest of the plumage is often poorly preserved, mainly as a dark ‘halo’ round the neck, body and wings ([20], [60] and [147] and pers. obs. on a dozen specimens). Even secondaries are difficult to distinguish, and there is no evidence of tertiaries, as shown in [97 (Fig. 6.2)]. The body feathers appear superficially rather similar to the ‘hair-like feathers present on many of the non-avialan theropods, or at least more simple than modern pennaceous body contour feathers, and this appears also to be the case with early enantiornithine birds, such as like Longipteryx [146], although the early ornithurine Archaeovolans is said to have contour feathers on the neck [16] and [57].

Thus, the body feathers of Archaeopteryx may not have been quite as modern as the remiges and rectrices could suggest, and as illustrated in most reconstructions, including [48] and [49]. They could have been simpler than in most modern birds, except perhaps flightless ratites such as Apteryx and Casuarius, with less developed and simpler vexillae (probably a secondary reversal). Because detailed studies of the variably preserved plumages of the Chinese feathered birds are still wanting (but see [16], [93], [145] and [147]), it does not follow that neither their body feathers nor those of Archaeopteryx were very different from modern contour feathers, especially as dromaeosaurs had some with a superficially rather modern appearance [81] and [141]. But it cannot be ruled out that the early birds had body plumage more like the simple body covering of most non-avialan theropods, that is unbranched ‘hair-like protofeathers’ [15], [93], [138] and [139], albeit sometimes with a little branching [16] and [25] and then called ‘down-like’ – and definite down-like structures were described in e.g. Protarchaeopteryx [93] and Scansoriopteryx [16] and are in Caudipteryx (NB’s pers. obs.).

The history of the preparation of the Berlin specimen can roughly be followed from figures in [26], [30], [87], [97] and [119], and the photos presented in this study, taken in 2002. What probably represents the first drawing of the Berlin Archaeopteryx from 1879 (Fig. 3) clearly illustrates the specimen with feathers around the legs, back and neck. Feduccia [36 (1999 edition, pp. 29 and 113)] argues that this early drawing shows true body feathers, which were subsequently removed during the ongoing preparation of the specimen, and the ‘feathers’ present on the slab today are thus merely preparation marks. These ‘feathers’ appear more slender and morphologically simpler than the remiges and rectrices. This is in contrast to the remains of the purported ‘feather trousers’ of the Maxberg specimen, which Heller [51 (pls. 5 and 15), 107 (Fig. 1b)] illustrates as near-typical contour feathers, i.e. pennaceous, presumably sempiplume feathers.

Dames’ [26] plate shows that in his day almost nothing was exposed of the pelvis, and little of the left femur, knee and proximal metatarsus. Ribs and gastralia were visible immediately in front of the right femur, but no feathers are seen in that area, apart from remiges from the elbow just anterior to the right knee. Posterior to the knee is an irregular lump of matrix, and the striations (described below) behind the right tibia reach the left knee region. Posterior to the left femur is a large patch of faint, slightly curved and near parallel striations, extending away from the femoral shaft at a near perpendicular angle. This is less evident in [49], but rather evident in all the photos by [88], [97] and [119] and the new photos presented here. Purported feathers along the posterior part of the back are evident in [26], but very faint in Heilmann’s [49] drawing. As noted below, they are clearly body feathers.

The left femur and knee, as well as ischium and pubis, had been disclosed, and the area behind the right knee had been evened out by preparation before Heilmann [49] made his drawing (about 1924). His illustration shows an area with very thin, sub-parallel, and very tightly set striations (probably preparation marks), nearly perpendicular to the distal femur. These striae seem to have vanished after a preparation done prior to the 1950s, as seen in de Beer [30] (note the illumination from opposite direction). Before the 1980s, only slight additional preparation around the distal pubis and between the knees was done, removing a little more of the feathers behind the right tibia (Fig. 1 and Fig. 5). Additional preparation around the skull has removed a little of the tips of the anterior feathers on the back (Fig. 5). Since the 1980s, only a little detailed preparation around the finger claws has been done, and a bit along the neck has made the slab paper thin, actually piercing several small holes through it (pers. obs., 1996, 2001).

The above is interesting since the current appearance of the Berlin Archaeopteryx, compared to the early drawing (Fig. 3, although erroneous in some respects; compare to Dames [26 (pl. 1)] and Heilmann [49 (Fig. 2)] who also show these imprints) clearly demonstrates that systematic removal of major portions of the purported feathers, as suggested by Feduccia [36], has not taken place. Either the faint striae at the base of the neck and the legs are feathers and still present, or they were never present at all.

In May 2001, we studied the slab and counterslab (Fig. 1 and Fig. 2) of the Berlin specimen of Archaeopteryx, using traditional light microscopy and camera lucida for illustrations. On the ventral side of the cervical vertebral column are faint, superficially ‘hair-like’ impressions, which may or may not be remains of feathers. The individual striations are faint and the matrix appears a little too coarse to identify these structures as evidence of feathers with any degree of certainty, although this appears a likely explanation.

Dames [26 (p. 39)] thought that the weak striations in front of the chest at the base of the neck (see also Fig. 10) had been made by wet and collapsed pennaceous body feathers, but his experiments with soaked chicks and doves could not reproduce the pattern seen in Archaeopteryx. In fact, he thought that these feathers might have formed a ‘collar’ like in some vultures because of the way they were bent (his pl. I). Other scholars accepted their presence [30] and [50], seemingly at face value. These earlier authors did not doubt their feathery nature, while most others simply ignored them, but their simple, ‘hair-like’ morphology could be indicative of their having been rather simple or a preservation artefact (see further below).

Much more informative are structures on the slab around the dorsal vertebral column and left and right tibiae. The impressions above the dorsal vertebral column are clearly feathers. Historically there seems not to have been much dispute about this, although they are not usually included in discussions of the Archaeopteryx plumage. The impressions could not have come from the tail region as they extend considerably past the ilia. The remiges and rectrices are preserved in what appear to be approximately their natural positions, and none appears to have come away and settled elsewhere, which would have had to be the case if the feathers along the back were indeed rectrices. Furthermore, the impressions radiate away at near right angles from the dorsal vertebral column, and are not jumbled, but arranged in what appears to be one row along the vertebral column, again strongly suggesting that their occurrence along the back was not a result of post-mortem displacement.

The dorsal feathers extend from slightly posterior to the ilia to the middle of the dorsal column, after which they seem to fade out (Fig. 5). There are about eight of them with nearly parallel raches. The distal tips of the anterior ones were removed by the preparation of the skull roof. But the anterior feathers were probably slightly shorter than the posterior ones, which when bent backwards could well have covered the proximal, muscular part of the tail. Anteriorly, where the feathers fade out, are some irregularities in the matrix (and a bone?), and of course the skull. It is impossible to verify if there was originally only a single row near the dorsal midline, and one (or a few) could well be hidden below the thin laminae of limestone. The reconstruction in [30] shows rather large contour feathers in about two rows in this region.

Their feathery nature seems unequivocal, as they bear resemblance to undisputed rectrices, but they appear to have been smaller and less rigid. Easily discernible are central raches, faint and sometimes incomplete and(or) overlapping outlines of the apparently symmetrical vexillae, and in several places even clear impressions of barbs. Barbules cannot be made out, but the ordered appearance of the vexillae would suggest the presence in life even of modified barbicles (hamuli). Toward the apices, the vexillae lose their ordered structure and appear more ‘plumulaceous’, suggesting that hamuli were absent in this region. This further corroborates the notion that they are not rectrices, because rectricial vexillae appear to preserve their three-dimensional integrity all the way to the apices. Thus, the dorsal feathers alone would suffice in assigning evidence of body contour feathers to Archaeopteryx, and clearly pennaceous ones at that.

Faint impressions are also evident posteriorly and perpendicular to the left femur (Fig. 7), but contrary to popular opinion this does not represent the ‘hind wing’ of Beebe [4], who believed that the feathers along the back represented the pelvic alar wing (pp. 49–50) that had drifted away from its original position (see also Fig. 5). The femoral impressions are similar to the impressions around the tibiae, but bend slightly away from the left tibia. The impressions around the left tibia could stem from the tail feathers, since the left leg lies adjacent to the tail, but the orientation of the striations in the impression, bending towards the tibia, is different and the microstructure also differs from that of undisputed rectrices, notably in the weak shaft impressions and lack of clearly discernible vanes and barbs (see further below). However, the patch around the posterior part of the left tibia appears very similar to those posterior and anterior to the right tibia. Thus, the latter were the primary subjects of this study. There are a substantial number of reasons as to why these impressions are not man-made preparation marks, as has been suggested [36], but are truly feathery in origin. Some of the striations proximally behind the left femur may, in fact, be imprints of feathers that hafted to the thigh. This is likely partly to have prompted both [4] and [33] to suggest ‘wings’ on the hind limbs (see de Beer’s [30] remarks, and further on Tetrapteryx below).

A comparison with preparation marks from around the slab reveals a number of substantial differences between the presumed leg feathers and undisputed preparation marks. The patch anterior to the right tibia (Fig 6; henceforth termed the integumentary patch) is subdivided into a number of near-parallel fine and very faint striations, which, although often fragmentary and with sections missing or very indistinct, clearly extend all the way through the patch in nearly straight lines. The striations are set at roughly 40^o to the long axis of the tibia. They are clearly not the result of smaller individual lines adjacent or in succession to each other. The striations are up to 3.5 cm long, but the actual length is tentative because they are very faint and often not preserved in their entity. The striations do not reach the tibial bone but terminate a few millimetres anterior to it, just like the integumentary structures preserved with the Chinese non-avian and avian theropods. This is probably because these structures were imbedded in the skin, which has long since rotted away.

There are several reasons why these are not preparation marks. First, the striations in the integumentary patches are uniformly fainter than preparation marks (e.g., along the long axes of the bones, feet, hands, claws, vertebral column, adjacent to remiges, etc.). Second, they are much longer, forming lines up to 3.5 cm in length. This is up to 10 times the length of typical preparation marks around the feet. Third, preparation lines are not only shorter, but also much more clear in outline. Each preparation mark is a short, sharp longitudinal excavation with well-defined edges that are offset clearly from the surrounding matrix. In a typical preparation area are many such small, sharp lines lying right next to and in succession to each other. The striations in the integumentary patch are not broken up like this.

Fourth, when examining a larger area that has been prepared (e.g., along a long bone or around the feet), it is evident that preparation marks that run parallel to each other are found in localized patches within the preparation area. This is due to the preparator’s work from different angles in order to remove matrix most efficiently and with less danger of damage to the fossil (compare, e.g., descriptions of preparation given by [38], [126], [135] and [136]). This creates a pattern of many individual, small, sharp, clearly offset lines lying parallel and in succession to each other, and right next to this localized patch is another with the exact same morphology within the patch (i.e. all individual lines are parallel and/or in succession). The angles of the individual lines in this patch are, however, different from those of the previous patch, since the preparator has now moved the slab slightly. This differs markedly from the long, fine striations inside the integumentary patch.

Fifth, preparation marks are not only sharper and more clearly offset from the surrounding matrix, but are clearly also deeper. Sixth, the striations in the integumentary patch are up to 3.5 cm long. Had the patch indeed been the result of preparation, it seems hard to explain why 3.5 cm of matrix should have been taken away, since this is much more than has been taken away anywhere else on the fossil. This seems particularly unlikely, since matrix has clearly not been removed at all in this area! This is easily recognized since the matrix lies nearly halfway up the mediolateral extent of the tibia.

Comparison of the Berlin specimen with older drawings reveals that preparation has taken place around the pelvic area, mainly to free the pubes, thus slightly damaging the integumentary patch behind the right tibia, whereas the integumentary patch in front of the right tibia is today as it was in 1877. In other places where the matrix has been prepared away from the bones, these are nearly completely exposed, and preparation has been done only on the main slab. Additionally, as noted above, the fine striations do not start until a few millimetres away from the bone. Why should a preparator prepare an area that does not include the bone itself?

The above differences are most clearly demonstrated on the small integumentary patch that lies behind the right tibia (with longitudinal striations running at a roughly 60^o to the long axis of the bone), where the fine striations have been damaged by preparation marks toward the distal part of the patch (Fig. 6). Under the microscope, the difference between the fine striations and undisputed preparation marks is striking. Some of this posterior patch has in fact been removed because of preparation around the knee and to free the pubes, and by preparation down the distal part of the tibia. Interestingly, towards the apex of the posterior integumentary patch, very fine, short lines can be made out that extend from the longitudinal striations. These small lines are set at roughly 80–85^o to the long axis of the striations, exactly as would be expected if the long striations represented soft raches and the smaller, finer lines represented barbs. A vexillum cannot, however, be traced, because the outline of each feather is quite blurred.

Seventh, even slight preparation clearly changes the colour and microstructure of the slab. This is easily recognized by comparing the surface structure of the integumentary patch to any preparation patch on the slab. The colour and surface structure of the integumentary patch does differ from parts on the slab where nothing is preserved (see, e.g., [14], [52] and [107] for explanations of local differences in sediment structure on the Archaeopteryx fossils), but these differences are clearly of a different nature to the changes brought about by preparation. Rather, the colour and surface texture of the integumentary patch is very similar to those patches with undisputed feather impressions, such as wings or tail.

Finally, examination of the counterslab (Fig. 2 and Fig. 9) reveals structures of the integumentary patches that are less discernible on the slab. The main slab contains nearly all the bone material, and the counterslab contains clear impressions of forelimbs, including the hands, legs, feet and vertebral column, including the tail. The central part of the fossil is, however, missing on the counterslab. Bone is present only as rib fragments and fragments of the caudals near the articulating facets of several vertebrae. Impressions of the remiges and rectrices are, however, excellent, and for many feathers a central rachis, both vexillae and their barbs can be made out with the naked eye.

Microscopy failed to identify with certainty any evidence of barbules, because the otherwise very fine sediment simply appears too coarse for the preservation of such delicate structures, although the ordered structure of the vexillae strongly suggests the presence of barbicles modified into hamuli, as noted above. In all specimens of Archaeopteryx, we have studied (four skeletons and the isolated feather) possible barbules that can only be made out in a few regions of the distal part of the slab of the isolated feather. The isolated feather was clearly preserved under different circumstances from the skeletal specimens [47].

The integumentary patch of the right tibia on the counterslab (Fig. 8) bears the same fine striations as are preserved in the patches around the anterior and posterior face of the tibiae on the slab, especially the right tibia. On the counterslab it is, however, evident that the fine, nearly 3-cm-long striations in several places have short structures preserved that extend away from the primary striations at a roughly 70^o (Fig. 9). These resemble the barbs of the remiges and rectrices (provided that the shallow longitudinal striations are indeed imprints of rachides), though they appear to be much shorter. Since no preparation has been done at all on this part of the counterslab, none of the above structures can be preparation marks.

Overall, the patches around the legs (presumable also the femur) are so different from preparation marks that this explanation must be discarded. The most likely explanation is that they represent the remains of body contour feathers, as observed originally by both Evans [33] and Dames [30], and where barbs cannot be seen, they are still covered by a very thin limestone crust, blurring the outline of the structures.

The integumentary patches on the slab failed to show evidence of a distinct vexillum or tertiary structures (barbules). On the counterslab, secondary structures (barbs), although clearly present, are not preserved in very many places and not very distinct (Fig. 9). However, faint vexillae may be reconstructed from several of the long striations (evidently the raches) by following the small, short striations until their termination (Fig. 8). Our illustrations of the Berlin specimen bear some resemblance to those of Heller [51 (pls. 5 and 15)] for the Maxberg specimen, albeit with proportionally more narrow vanes with more squarish tips and more simple construction of the vanes, and not so fully modern in appearance (cf. Fig. 9 and Fig. 10).

Some early authors (e.g., [26] and [125]) inferred that the tibia feathers in the Berlin specimen formed ‘feather trousers’, whereas Evans [33] suggested that they were more confined to the periphery of the bone, forming some sort of vane, and being directly involved in flight. The most important conclusion from the above observations is that some of the striations (the proximal ones behind left femur) may represent long hind-limb feathers, and this seems to be what prompted Beebe [4] to accept from a photo of the Berlin specimen “a diminutive wing” in the pelvic region (contra [99]), as discussed by Heilmann [49 (pp. 194–196)], who thought that Archaeornis’ “thigh feathers are […] too weak” for a wing. Heilmann does not discuss these ‘feathers’ of the fossil any further, but he does reconstruct his Proavis with a border of short feathers behind the thigh (Figs. 139 and 142). He does not show the thigh in the reconstructions of ‘Archaeornis’, but Beebe’s rather schematic figure clearly shows them (but not much on the lower leg).

The presumed tibial and femoral feathers of the Berlin Archaeopteryx appear to have assumed an important role, first in Evans’ [33] introductory remarks about flight with both fore- and hind limbs, and later in Beebe’s [4] Tetrapteryx hypothesis, which was discussed in detail by Heilmann [49 (194 ff.)] with some part of Beebe’s idea about this intermediate stage being accepted (see further above). De Beer [30] and Heller [51] accepted these feathers around the legs (though not as ‘wings’, but rather as ‘breeches’ – note that de Beer seems to confuse femoral feathers and breeches on the lower legs).

Recent finds of some small and primitive dromaeosaurs show long feathers behind the legs [81] and [141], and Microraptor possibly even had a ‘wing’ of very elongate contour feathers behind the whole length of the leg, including the metatarsals carrying the longest and asymmetrical ‘flight feathers’ [142] (but see [91]). This has revived the ‘Tetrapteryx-theory’, and prompted Xu et al. [142] to implicate such ‘hind wings’ in their common ancestor with birds (that is, in the stem-Paraves in the sense of Sereno 1999 [110]), but later lost in the bird-lineage. Prum [99], however, warns that perhaps the ‘hind wings’ were apomorphic to dromaeosaurids, implying that the direct lineage towards birds never had such hind wings, and urged that Archaeopteryx be closely inspected to illuminate this problem, as we have done here.

We conclude that Archaeopteryx had rather long contour feathers along the legs, apparently both upper and lower, and that they may have formed ‘edges’ rather than ‘trousers’. Thus, they might be considered remnants of former ‘hind wings’ in the bird-lineage, but only finds of feathered representatives of outgroups close to Paraves may solve this puzzle [10] and [22], e.g., better preserved specimens of Protarchaeopteryx (cf. [61]).

The body contour feathers on the back and legs of Archaeopteryx appear to have been fairly modern in appearance, although perhaps not as similar to feathers on most extant, volant birds as are the remiges. However, this is tentative because their state of preservation is not good enough to draw firm conclusions about their detailed morphology (contra [26]). Both the feathers that extend from the back and legs (Fig. 6, Fig. 5, Fig. 8 and Fig. 9) appear to have been elongate, with rather narrow vanes (i.e. not as wide, ovoid and plate-like as a ‘typical’ semiplume body contour feathers) and were clearly more pliable than the flight feathers. Furthermore, they appear not to have had vanes as closed as the flight feathers, as evidenced by the morphology of the feathers along the back (Fig. 6).

Evidently, they were not, however, nearly as simple and hair-like as the feathers recovered with the primitive coelurosaurian Sinosauropteryx [15] and [25]. More advanced theropods, such as Sinornithosaurus [139] and [141] and Microraptor [142] had more complex, albeit still superficially ‘hair-like’ body contour feathers and down-like structures, corresponding approximately to stage 2 of Prum’s [99] model for feather evolution. Protarchaeopteryx is said to have ‘down’ and apparently had pennaceous contour feathers [61] and [93]. However, these were recovered separately on the slab, not associated with any particular part of the body. Accordingly, their identification as body contour feathers should be regarded as tentative, and they may represent coverts instead. Caudipteryx also has remiges and rectrices and even downs (pers. obs. and [93]); as noted above (Fig. 4), they might correspond to Prum’s [99] stage 2.

In Cryptovolans [81] and Microraptor, there appears to have been body contour feathers of a superficially modern appearance, those of Microraptor being called ‘plumulaceous’ [142]. Whether these correspond to Prum’s [99] stages 3a (rachis + unbranched barbs) or the more complex stage 3a+b (rachis, barbs and barbules), is still not clear, and will require more analyses of vane morphology. The body contour feathers in Archaeopteryx also appear to correspond to Prum’s [99] stage 3, but the stage of preservation is such that even careful analyses failed to distinguish between stages 3a or 3a+b. At an approximate length of 3.5 cm along the epipodials this may seem a lot for a bird of Archaeopteryx’ size [144] and could represent the remains of ‘hind-limb wings’, as are present in Microraptor gui [142] (but see [91]) and possibly M. zhaoianus [140]. In Archaeopteryx the ‘hind-wings’ were apparently much less prominent than in M. gui, and probably absent from the metapodials, thus perhaps representing a more modern way of flight with the forelimbs only. Thus, if Archaeopteryx really did possess a vestigial ‘hind-limb wing’, this might be regarded as an avialean plesiomorphy.

The plumages of the many specimens of confuciusornithid birds recovered from northeastern China are usually in a much poorer state of preservation, and are likewise preserved mainly as carbonised residue, not as imprints [20], rendering analyses of their detailed morphology difficult. At a glance the body contour feathers appear hair-like, not unlike the simple filaments recovered with several of the non-avian theropods. However, the phylogenetic position of the Confuciusornithidae (Fig. 4) and the probable morphology of the body feathers, at least around the legs and back, of Archaeopteryx (and feathers in some dromaeosaurs), suggest that future analyses of the detailed morphology of body feathers in well-preserved confuciusornithids will reveal pennaceous, not simple and ‘hair-like’ body feathers.

The Berlin Archaeopteryx exhibits well-preserved rectrices and remiges with coverts [106] as sharp imprints, but rather weak imprints of the feathers on the back and hind limbs, and questionable structures below the throat and on the chest. Thus, two models for the type of preservation must be taken into account.

The pattern observed today approximates a realistic image of the plumage in the living animals, although a few layers of feathers may be hidden under thin crusts of limestone (see [106] and [107]). The near-absence of larger feathers on the body (apart from the back) is primary, indicating that, at most, rather simple ‘down-like’ feathers or ‘hair-like’ cryptoptiles (e.g. not unlike those of Sinosauropteryx [25]) covered the body. These were lost during fossilization. The faint striations below the neck and at the chest could be the only remains of such feathers. Rietschel [106] discussed the preservation of the wing and tail feathers as imprints of only the ventral surface, but the body feathers are not even mentioned, nor were they discussed by Stephan [119].

Alternatively the observed pattern is only the trivial result of taphonomic processes (cf. [28] and [29]; criticised by [63]) leaving only reasonably well-preserved feathers as a ‘halo’ at the circumference of the carcass (already suggested by [90] and [109]), indicating that the body itself may well have been covered by contour feathers at least as well developed as the ones along the back. The preserved pattern is exclusively secondary, and, thus, of limited significance. Such an explanation has been offered for the preservation of feathers in Confuciusornis [20] and in the Middle Eocene deposit of Messel [53] and [72]. Model (2) is the less interesting alternative and is close to an ad hoc assumption. There is little reason to believe that Archaeopteryx did not have some feather covering all over the body, apart from perhaps the feet and parts of the face. Both models can be tested by new and better-preserved fossil finds.

How may the feathers have been arranged? As [119] noted, there is universal agreement that the rectrices, symmetrically preserved in the London and Berlin specimens, were arranged in one row along the lateral sides of the tail, although in the former specimen the caudals are viewed from above and in the latter from the side. To our knowledge, it has never been argued that they were arranged vertically, or that rectrices were present all the way round the caudals (as appears to be the case with the strange fibres recovered with a Psittacosaurus specimen [73]), and the cryptoptiles in Sinosauropteryx [15] and [25]. Accordingly, an ad hoc explanation of missing rectrices owing to a decomposing carcass has not been advanced in this respect.

With the feathers preserved rather symmetrical along the lower legs (also a slender structure with little flesh for decomposition) the situation is, however, different. Most often these have been regarded as forming ‘feather trousers’ or at least having had a circumferential distribution, analogous, or most probably homologous, to extant neornithines. It cannot, however, be ruled out that they too were primarily arranged as feather borders, at most a few rows in thickness. The extremely large hind limb feathers in Microraptor are interpreted in just this fashion [99] and [142], and not as the leftover remains of a preservation artefact. Additionally, it would be a rather peculiar incidence if just one (or a few at most) of the feather rows got preserved, lying with near-parallel raches in apparently just one level, without any trace of neighbouring rows, if indeed they were present. Yet such seems to be the case in Archaeopteryx (and Microraptor). We could only determine one incident of a crossing of two supposed raches.

Accordingly, an interpretation of Archaeopteryx as endowed with a sort of ‘Tetrapteryx - hind limb wing’, albeit reduced compared to both Evans’ [33] and Beebe’s [4] hypothetical Proavis-stages (and certain members of the plesiomorphic sister taxon Dromaeosauridae). Minimally the preservation pattern appears to support “feather fringes” along the legs of Archaeopteryx instead of ‘feather trousers’.

If model (1) for the preservation of the plumage of the Berlin Archaeopteryx is correct, this implies that some parts of the body (back and legs) bore, at most, a few rows of larger feathers, while other areas (main part of the corpus, perhaps most of the neck) had, at most, smaller proto-feathers or ‘proto-down’. This implies that some areas had large pennaceous feathers while others lacked them and only had ‘simpler’ structures, and perhaps even were unfeathered. Thus, something similar to pterylae could have been present on the legs and at the midline of the back, i.e. ‘proto-pterylae’ and ‘proto-apteria’. Obviously, this idea can be tested by new discoveries of fossil of either Archaeopteryx or close relatives, with more well-preserved plumage.

In the development of pterylae in extant birds (even comprising the down plumage of more simplified structures [11] and [66]), it becomes evident that the middle of the back and legs (both upper and lower) nearly always carry pterylae with pennaceous feathers. There are many other areas with pterylae and apteria in general in neognaths, save penguins that have a rather uniform feather covering. Such uniformity is also found in ratites but not in tinamous, implying an uncertainty of the polarity of this character in the last common ancestor of crown-clade Aves (= Neornithes). Did this common ancestor have apteria or a uniform feather covering?

The pattern of pterylae (‘pterylosis’ according to [66 (p. 73)]) is generally also mirrored in the pattern of papillae for the embryonic down (neoptiles – see [11]). The stripes on the back are often paired, but set close to the midline (op. cit., Fig. 50), or unpaired in the anterior region of the back, and these tracts are composed of only a few rows of feathers. As mentioned above, in the Berlin specimen, we cannot be sure whether there was only a single row of feathers as observed, or one or a few rows hidden below the laminae. From a functional point of view, this one (or a few) row(s) along the back hardly stood upright in life, as in the fossil. It seems more likely that the feathers were lying down to cover the muscular part of the proximal tail. The row as preserved apparently ends around the middle of the back, at least it fades out in the matrix. Whether this was the case in the living animal is unknown, but smaller and simpler feathers may well have covered the anterior part of the back and neck. There is no obvious, taphonomic reason why larger feathers should have disappeared from these areas when being preserved posteriorly.

It is perhaps of interest that in some birds the ontogenetically earliest feathers (neoptiles) to develop are often of considerable length, and are situated at the midline above the pelvis and the posterior back and on the thighs [11], e.g., in crows, and in others species, e.g., herons, these long feathers are also present on the thighs, and are arranged in double rows close to the entire midline of the back (op. cit., Fig. 26). Could such a pattern be yet another case of recapitulation or is it just fortuitous? We prefer the former, which supports Haeckel’s theories (see also [7]). Thus, it is possible that the early development of these long neoptiles in extant birds, to be replaced by pennaceous feathers in the pterylae, could represent the pattern observed in archaic avialeans, as exemplified by Archaeopteryx.

Such ‘proto-pterylosis’ may in fact be even older, characterising the ancestral maniraptoran, as indicated by the plumage in Caudipteryx and dromaeosaurs. Thus, the polarity within primitive Neornithes may be solved. They had pterylosis, not a uniform feather covering. On the other hand, we believe that the ‘patterning’ of the integumental covering in Sinosauropteryx as shown by the ringed appearance of the tail [15] and [25] is a preservation artefact conditioned by the vertebral arches (pers. obs.). Sinosauropteryx probably had a more uniform covering of cryptoptiles [25]. Accordingly, early coelurosaurs may not yet have been ‘pterylized’.

The Berlin specimen of Archaeopteryx is the only one to preserve remnants of body plumage (apart from the now missing Maxberg specimen), which is here described in detail. The London specimen shows no indications of body feathers. Most of the body feathers that were originally present are still present today and preparation has only damaged a few. They have been largely overlooked in 130 years of debate on Archaeopteryx, feathers.

3–4-cm long feathers are present around the legs, and although not well preserved, shafts and in some places parallel, closely set barbs can be observed. Barbules are not visible, but may be inferred. The vanes can be vaguely outlined, and the feathers may have formed ‘trousers’, or, alternatively, narrow ‘edges’ on the fore and aft sides of the limbs. Less well preserved feathers of roughly the same type are present behind the femora, albeit only as weak impressions of parallel raches. Undisputed feathers with raches and faint vexillae are present nearly perpendicular to the vertebral column, from the ilium to halfway up the back. Their slightly bent shafts appear quite similar to the feathers on the lower legs. The dorsal row of feathers (one row apparently, although more may be hidden below thin limestone crust) probably covered the back and the proximal tail in life. These feathers are clearly body contour feathers.

Faint imprints in front of the breast may be shafts of more simple and ‘hair-like’ feathers, or collapsed pennaceous contour feathers. No other feathers can be observed on the body, and if this represents the original distribution, then the pattern of feathers along the back and legs and either none or only small, simple and less easily preservable feathers on the rest of the body could indicate the presence of pteria and apteria and some sort of pterylosis (here termed ‘protopterylosis’).

Comparison with feathered theropod dinosaurs and primitive birds lead to the following conclusions: (1) feathers, even pennate remiges and rectrices, originated long before flight, no later than in stem-Maniraptora, which also had down; (2) simple hair-like proto-feathers with little branching (resembling the hypothetical ‘cryptoptiles’ of ornithological literature) are even older, a synapomorphy of at least Coelurosauria; they might have been retained on part of the body up to the level of confuciusornithids or even early ornithurines and enantiornithines; (3) ‘proto-pterylosis’ is characteristic of Maniraptora, as some regions have remiges and rectrices, others have pennate body contour feathers, and others have down or ‘cryptoptiles’, the latter two types most likely covering ‘proto-apteria’; (4) the feather rows on the back and the thigh of Archaeopteryx may correspond to the rows of very long neoptiles present in the pteria very early in the ontogeny of several living birds. If so, they perhaps illustrate a case of Haeckel’s ‘biogenetic law’; (5) comparison of the ‘feather trousers’ of Archaeopteryx to the so-called ‘hind wings’ of primitive dromaeosaurs indicates that their common ancestor minimally had long feathers on the legs, not necessarily ‘wings’, so whether birds went through a ‘Tetrapteryx stage’ remains uncertain.