The fossil specimens are deposited in the collections of Senckenberg Research Institute Frankfurt, Germany (SMF), Hessisches Landesmuseum Darmstadt, Germany (HLMD), and Generaldirektion Kulturelles Erbe Rheinland-Pfalz, Direktion Landesarchäologie, Referat Erdgeschichte, Mainz, Germany (PW). Osteological terminology follows Baumel and Witmer (1993).

A phylogenetic analysis was performed based on the character matrix of Mayr (2013), with the same settings as in this latter study. The following six character scorings were added or modified: character 2 (bill approximately one-half of total skull length): scored as present (0) for M. brevidactylus; character 4 (eminentia articularis of quadrate): scored as absent (0) for M. brevidactylus and Sandcoleus copiosus; character 5 (mandible with processus retroarticularis blade-like and elongated to approximately one-sixth of the skull length): scored as present (1) for M. brevidactylus; character 26 (configuration of foramina vascularia proximalia): scored as (0) for M. brevidactylus (two foramina present) and as (2) for Selmes absurdipes (single foramen on lateral side of shaft).

COLIIFORMES Murie, 1872

COLIIDAE Swainson, 1837

Masillacolius Mayr and Peters, 1998

Masillacolius brevidactylus Mayr and Peters, 1998

Referred specimen. SMF-ME 11322 (skeleton lacking the upper beak, the right hand, and the left tarsometatarsus; Fig. 1).

Locality and horizon. Messel near Darmstadt, Germany; Early Eocene (Lenz et al., 2015).

Measurements (left/right, in mm). Skull, estimated length, ∼39; humerus, 20.4/20.3; ulna, ∼20.4/–; carpometacarpus, ∼12.9/–; femur, 22.4/22.8; tibiotarsus, 30.2/29.1; tarsometatarsus, –/19.8.

Description and comparison. In the following, only features that were not already previously described are mentioned (Mayr and Peters, 1998). The upper beak is not completely preserved, and the tips of lower jaws are likewise broken. However, the proximal portion of the right ramus mandibulae is well visible and allows the recognition of a long, blade-like processus retroarticularis (Fig. 2A, B). The ramus mandibulae is of moderate dorsoventral depth, and unlike in crown group Coliiformes, a fenestra mandibulae is absent.

The caudal portion of the processus mandibularis of the quadrate (Fig. 2B) forms a marked shelf-like lip as in extant mousebirds. Unlike in living mousebirds and Oligocolius (Mayr, 2013), however, an eminentia articularis is absent and further unlike in extant Coliiformes, the processus lateralis does not form a marked, dorsally directed process.

There are 18 praesacral vertebrae as in extant mousebirds (most other neornithine birds have 19 or more praesacral vertebrae). The cervical vertebrae correspond well with those of extant Coliiformes (Fig. 2C, D). As in the latter, only the third cervical vertebra departs in its shape from that of the caudally following ones, which have much deeper lacunae interzygapophysiales. In most other extant birds, both the third and fourth cervical vertebrae are characterized by the absence of well-defined lacunae interzygapophysiales and have a more rectangular outline than the following cervical vertebrae. Whether the fourth to eighth vertebrae exhibit a bony strut connecting the processus transversus with the corpus vertebrae, as in extant mousebirds, cannot be discerned in the fossil.

The extremitas omalis of the coracoid has a similar shape to that of Chascacocolius and the facies articularis clavicularis overhangs the sulcus supracoracoideus (Fig. 3B). A foramen nervi supracoracoidei is absent (this foramen is present in the Sandcoleidae). The facies articularis scapularis merges into the facies articularis humeralis. Being short but wide, the acromion of the scapula has a squarish shape and the scapular corpus is straight and narrow (Fig. 3B). Unlike in crown group Coliiformes, the extremitas omalis of the furcula does not form a well-developed processus acromialis.

The humerus of Masillacolius is short and stout. A notable and previously unreported feature is the presence of a well-defined, dorsally directed tuberculum supracondylare dorsale (Fig. 3C), which is not well visible in previously described Masillacolius specimens. The development of this tubercle varies among extant Coliiformes, being equally well-defined in Urocolius (Fig. 3D), but more weakly developed in Colius. Among Eocene stem group Coliiformes, its occurrence is also variable, and a small tuberculum supracondylare dorsale is present in Chascacocolius but absent in Selmes, Celericolius, and Primocolius. On the right radius, a large nutrient foramen is visible in the proximal third of the bone.

In the new specimen, the proximal end of the tibiotarsus is not covered by overlying bones. This reveals that the length measurements given by Mayr and Peters (1998) are too low, owing to the facts that the proximal ends of the bone are hidden in the holotype of M. brevidactylus and that the tibiotarsus is broken in the second specimen described by Mayr and Peters (1998). In the holotype and the new specimen, the distal end of the bone is of equal width throughout and not widened on the level of the condyles as in extant Coliiformes and most of other neornithine birds.

As already detailed by Mayr and Peters (1998), the second, third, and fourth toes of Masillacolius have abbreviated proximal phalanges, whereas in extant mousebirds only the proximal phalanges of the fourth toe are shortened. The ungual phalanges are deeper than in extant mousebirds and other stem group Coliiformes. As in both feet of the holotype (Fig. 3A), all toes of the new specimen are detached from the left foot but are still closely attached to each other, which strengthens the assumption that Masillacolius had fully pamprodactyl feet. In the holotype there are sesamoid ossicles on top of the distal ends of the penultimate phalanges of each of the toes (Mayr and Peters, 1998: text – fig. 4), but these cannot be discerned in the new specimen.

Underneath the left foot a single large seed is preserved (Fig. 1 and Fig. 3G, H), which measures 6.7 × 4.4 mm and is tentatively identified as belonging to the Vitaceae. Unidentifiable organic matter is also preserved in the area of the gizzard, and because an accidental association of an isolated seed with a bird skeleton is highly unlikely in the Messel oil shale, where both, birds and isolated seeds, are not all too common, the seed is likely to represent dislocated stomach or gut content. There is also a small cluster of fine grit at the base of the (missing) upper beak, which may have been ingested by the bird.

Assignment of the new fossil to Masillacolius is straightforward and supported by the dimensions and proportions of the bones (see above concerning the length of the tibiotarsus), and by the derived morphology of the legs, with the tarsometatarsus being unusually long and the short toes bearing very deep claws. The new Masillacolius specimen is of significance, because it informs about the skull morphology of a poorly known stem group coliiform and therefore potentially adds to a better understanding of both, the phylogenetic interrelationships of early stem group Coliiformes and the evolutionary history of feeding adaptations in mousebirds.

It has recently been hypothesized that frugivory in birds did not play a major role in the early (that is, pre-Oligocene) stages of seed dispersal of angiosperm plants (Eriksson, in press). However, and as also noted by Eriksson (in press), stomach or crop contents containing seeds are already known from various avian taxa from the Early Cretaceous Chinese Jehol Biota (e.g., Zheng et al., 2011 and Zhou and Zhang, 2002). There exists a possibility that these early birds were granivorous rather than frugivorous, but the fact that the ingested seeds usually are uncrushed is more indicative of frugivory, and hence a dispersal function of the birds that ate them.

Stomach contents consisting of dense packages of seeds of various angiosperm taxa are preserved in several bird fossils from the Early Eocene of Messel (pers. obs.), and they occur in phylogenetically disparate taxa, such as zygodactylids and stem group rollers, with the extant representatives of the latter being carnivorous (Mayr, 2009). Eriksson (in press) noted that “Coliiformes was particularly diverse during the Eocene (…) but no evidence indicates whether they were frugivorous”. This is, however, certainly not true, and various Early Eocene stem group Coliiformes are preserved with seeds as stomach contents, including the sandcoleid Eoglaucidium (Mayr and Peters, 1998), as well as Selmes (Mayr and Peters, 1998 and Peters, 1999), and Masillacolius (this study, although, as noted above, the single seed is dislocated and not preserved in the area of the stomach or gut).

The highly specialized digestive system of crown group Coliiformes (Lauterbach and Prinzinger, 1994) also documents a long evolutionary history of frugivory of mousebirds. Still, however, there must have been substantial differences between the feeding ecologies of Eocene stem group Coliiformes and their extant relatives, as evidenced by the highly derived specializations of some of the fossil taxa.

The new Masillacolius fossil adds another taxon to the growing number of stem group Coliiformes with greatly elongated, blade-like retroarticular processes of the mandible. Such processes are now known from the Early Eocene Chascacocolius oscitans, Chascacocolius cacicirostris, and M. brevidactylus, and from the Late Oligocene Oligocolius brevitarsus (Houde and Olson, 1992, Mayr, 2005 and Mayr, 2013; this study; note that the asterisk symbol designating taxa with known skulls in Mayr, 2013: fig. 5 should have been assigned to Celericolius acriala and not to M. brevidactylus; the scale bar in Mayr, 2013: fig. 1 furthermore denotes 10 mm, and not 5 mm as erroneously stated). The mandible of the sandcoleid Sandcoleus definitively lacks elongated retroarticular processes (Houde and Olson, 1992: pl. 1d), and such were also reported absent in the Early Eocene Selmes (Mayr, 2001). The condition in Celericolius cannot be assessed, because of the poor preservation of the proximal end of the mandible of the holotype and only known specimen (Ksepka and Clarke, 2010); the skulls of Primocolius and Palaeospiza are unknown.

Retroarticular processes increase the leverage of the muscles that depress the mandible, and they represent a gaping adaptation for opening of the beak within substrate (e.g., Mayr, 2005 and Mayr, 2013). Among extant birds, such processes are most often found in species that forage for insects in soil or tree crevices, which are forcefully spread open, once the long and narrow beak has been wedged into. Possibly, therefore, the diet of Eocene stem group Coliiformes with long retroarticular processes contained a larger percentage of animals, although the occurrence of long retroarticular processes in the short-beak Oligocolius suggests that these processes also served other purposes (Mayr, 2013).

Blade-like retroarticular processes were previously considered an apomorphy of Chascacocolius (Ksepka and Clarke, 2010 and Mayr, 2005). However, their recognition in Oligocolius already raised the question whether these processes may be plesiomorphic for a coliiform subclade, which also includes the crown group representatives (Mayr, 2013). Previous analyses congruently recovered Selmes as the sister taxon of a clade including all other non-sandcoleid Coliiformes (Ksepka and Clarke, 2010, Mayr, 2013, Mayr and Mourer-Chauviré, 2004 and Zelenkov and Dyke, 2008). The analyses of Zelenkov and Dyke (2008) and Ksepka and Clarke (2010) further suggested that Chascacocolius branches next, and that Masillacolius and all other more modern-type Coliiformes form a clade. In the analysis of Mayr (2013), by contrast, the interrelationships of Chascacocolius and Masillacolius were poorly resolved, and Chascacocolius, Masillacolius, and Celericolius were recovered in a polytomy at the base of a clade, which includes all coliiform taxa to the exclusion of sandcoleids and Selmes. A re-analysis of the slightly emended data set of this latter study resulted in 30 most parsimonious trees (L = 63, CI = 0.63, RI = 0.75), with the strict consensus and majority rule consensus trees showing identical topologies to those found by Mayr (2013) (Fig. 4).

If Selmes, Chascacocolius, and Masillacolius are indeed successive sister taxa of a clade including more modern-type Coliiformes (Ksepka and Clarke, 2010 and Zelenkov and Dyke, 2008), elongated, blade-like retroarticular processes are likely to be an apomorphy of a clade including Chascacocolius, Masillacolius, Oligocolius, and crown group Coliiformes to the exclusion of Selmes and sandcoleids. Long retroarticular processes must then have secondarily become shortened in the lineage leading to the crown group, a scenario that gains some support from the fact that extant Coliiformes exhibit incipient retroarticular processes, which may represent vestiges of longer ones (Fig. 2E). Recognition of a wider distribution of greatly elongated retroarticular processes in stem group Coliiformes would potentially bear on the identification of C. cacicirostris, whose assignment to the taxon Chascacocolius was largely based on the very presence of these processes.

Ksepka and Clarke (2010) identified a single character, a straight humerus shaft, as an apomorphy of a clade including all coliiform birds to the exclusion of Sandcoleidae, Selmes, and Chascacocolius. This character is at least ambiguous, because the curvature of the humerus shaft is difficult to assess in taxa, of which only flattened skeletons on slabs are known (e.g., Masillacolius, Palaeospiza, Oligocolius). Moreover, the humerus of Primocolius actually shows the same degree of curvature as that of Chascacocolius (compare Mourer-Chauviré, 1988: pl. 1, figs. 7, 8 with Houde and Olson, 1992: pl. IIh–k). An alternative hypothesis to be considered is that Masillacolius and Chascacocolius are sister groups, with blade-like retroarticular processes being an apomorphy of a clade including both taxa and having evolved independently in Oligocolius. Unfortunately, the tarsometatarsus of Chascacocolius is unknown, with this bone being one of the most characteristic skeletal elements of Masillacolius. However, a decision for one of the two conflicting phylogenetic hypotheses outlined above would also be fostered if the presence or absence of elongated retroarticular processes could be verified in other Coliiformes, of which the skull is currently unknown.

In any case, Eocene stem group Coliiformes had proportionally longer beaks than their extant relatives. An obvious assumption is that the beak of mousebirds became shorter and more robust with increasing frugivory. This hypothesis is in concordance with the fact that the Oligocene Oligocolius, which already had such a short beak, appears to have been a specialized frugivore, as evidenced by the presence of a high amount of very large seeds that were ingested by one fossil individual (Mayr, 2013).

Other morphological attributes of stem group Coliiformes likewise indicate significant shifts in the ecomorphology of these birds. Coliiformes primitively had a derived foot morphology with shortened proximal phalanges of the second to fourth toes, which occurs in the Sandcoleidae and all stem group Coliiformes with known feet, except for Celericolius. In extant Coliiformes, by contrast, only the proximal phalanges of the fourth toe are shortened. Shortened pedal phalanges allow for a higher degree of curvature of the toes and thereby increase the grasping function of the foot. They are often found in birds that manipulate food items with their feet, such as owls, many diurnal birds of prey, and parrots. On the other hand, shortening of the proximal pedal phalanges also evolved in some groups to level the lengths of the second, third, and fourth toes, which usually constitutes an adaptation for clinging to vertical slopes and is found in, e.g., swifts. In stem group Coliiformes, both may have been the case.

Sandcoleids and Selmes probably used their feet for handling larger food and prey items, which is also suggested by the very long and pointed, raptor-like ungual phalanges of these birds. Masillacolius, by contrast, is characterized by a very long tarsometatarsus, which is longer than in all other Coliiformes. It furthermore appears to have had fully pamprodactyl feet with deep claws, and this unusual foot morphology may have served for clinging to vertical surfaces (Mayr and Peters, 1998). Swifts (Apodidae) are the only extant fully pamprodactyl avian taxon, and their feet are most similar to those of Masillacolius. In swifts, lengthening of the tarsometatarsus probably occurred in adaptation to a derived nesting behavior, which involves clinging to vertical surfaces (Mayr, 2015). Whether the peculiar foot morphology of Masillacolius represents a nesting or feeding adaptation cannot be said, but in analogy with swifts, a nesting adaptation is here considered more likely.