Eocene fossil is earliest evidence of flower-visiting by birds

Gerald Mayr , Volker Wilde


Birds are important pollinators, but the evolutionary history of ornithophily (bird pollination) is poorly known. Here, we report a skeleton of the avian taxon Pumiliornis from the middle Eocene of Messel in Germany with preserved stomach contents containing numerous pollen grains of an eudicotyledonous angiosperm. The skeletal morphology of Pumiliornis is in agreement with this bird having been a, presumably nectarivorous, flower-visitor. It represents the earliest and first direct fossil evidence of flower-visiting by birds and indicates a minimum age of 47 million years for the origin of bird–flower interactions. As Pumiliornis does not belong to any of the modern groups of flower-visiting birds, the origin of ornithophily in some angiosperm lineages may have predated that of their extant avian pollinators.

1. Introduction

Angiosperm flowers evolved characteristic pollination syndromes to attract insects and vertebrates as pollen vectors [13]. Although insects dominate in number and diversity, birds are important pollinators, especially in extant tropical and subtropical ecosystems [4,5]. The well-documented fossil record of flower-visiting insects goes back into the Cretaceous [1,6], but the onset of vertebrate pollination in general and that by birds in particular remains elusive.

Unless the phylogenetic position of a fossil bird is well constrained within a clade of known flower-visitors, it is difficult to prove nectarivory in the fossil record. Bill shape alone often does not provide unambiguous clues on the diet of a fossil bird, and stomach or intestinal tract contents are rarely preserved. If so, they usually consist of seeds or fish remains, which are already known from Mesozoic birds [7]. The earliest possible indication of plant pollination by birds comes from fossils of modern-type hummingbirds (Trochilidae) from the Early Oligocene of Europe [8]. A nectarivorous diet of these birds can, however, only be inferred from indirect evidence, such as the long beak and presumed hovering capabilities, and thus remains uncertain. Earlier, Middle Eocene, hummingbird stem group representatives still had a short and wide beak, which suggests that they were insectivorous like their apodiform sister taxa, swifts and tree-swifts [9].

Here, we report a bird skeleton from the Middle Eocene (47 Ma) of Messel in Germany with preserved stomach contents containing numerous pollen grains. This fossil constitutes the earliest and first direct fossil evidence of flower-visiting by birds and most likely predates the origin of extant avian nectarivores.

2. Material and methods

The fossil is deposited in the collection of the Senckenberg Research Institute, Frankfurt, Germany, with the collection number SMF-ME 11414a+b. Most of the skeleton is preserved on the main slab (SMF-ME 11414a), and the counterslab mainly consists of a resin cast of the impressions left by the fossil in the sediment. Because of the very small area with stomach contents, we could not sample the specimen for palynological processing. However, pollen morphology was studied by UV-induced fluorescence stereomicrography and low-vacuum SEM of the unsputtered fossil. High-resolution SEM information comes from a tiny sputtered sample. Pollen morphologies were assessed with online databases (http://www.paldat.org and http://pollen.usda.gov) and references [10,11].

3. Results

The new fossil is a complete skeleton and was found in 2012 (figure 1a). It belongs to Pumiliornis tessellatus, which is only known from two other specimens, neither of which preserves stomach contents [12]. The phylogenetic relationships of this very small bird are unresolved, with possible affinities to cuckoos (Cuculiformes) or parrots (Psittaciformes) having been discussed [12].

Figure 1.

Skeleton of P. tessellatus from the Middle Eocene of Messel (SMF-ME 11414a) with preserved stomach contents. (a) Overview of specimen, framed area indicates position of detail shown in (b), stomach contents are encircled. (c) Detail of stomach contents before drying of oil shale, with arrows pointing to dark, iridescent insect remains. (d) Skull, coated with ammonium chloride. (e) Right and (f,g) left foot in plantar view; the toes are numbered next to their ungual phalanges, bones in (e,g) were coated with ammonium chloride. at, accessory trochlea of fourth toe; ns, nostril; sh, sheath of ungual phalanges.

The new fossil shows excellent soft tissue preservation, and, in addition to the plumage, the sheaths of the pedal claws are preserved (figure 1f). Pollen is concentrated on the main slab in an area of about 4.9 × 2.6 mm next to the left femur, and in another, smaller, area left of the pelvis (figure 1). The position, size and sharply defined package of the grains suggest that they represent stomach rather than intestinal tract contents. Two pollen types of different size can be distinguished. The larger grains are spheroidal in equatorial view and have a diameter of 52–56 µm; their exact number cannot be counted as only part of the pollen is exposed, but the visible grains alone number several hundred. They are tricolpate–tectate with long and broad colpi almost reaching the poles (figure 2), the aperture membranes are covered with ectexinous granules, and the pollen surface shows small pores and tubercles. Some of the grains are still clumping together (figure 2c), but do not represent pollinia. A second, much rarer, pollen type is spherical, with a diameter of ca 13 µm (figure 2f); the apertures cannot be clearly characterized without destructive preparation.

Figure 2.

Details of pollen preserved in the stomach contents of P. tessellatus (SMF-ME 11414a). (a) UV-induced fluorescence stereomicrographs of main area with pollen and (b,c) selected pollen grains; framed areas in (a) indicate the position of the details in (b,c); the arrows in (b) point to the three colpi. SEM pictures of individual pollen grains; panel (d) is from a sample taken before preparation, and panels (e,f) are unsputtered details of the fossil slab. Except for the small pollen in (f), all pollen is of the large type.

The stomach contents of the new Pumiliornis specimen also contain a few unidentifiable insect remains (figure 1c), the number of which is however far too low for the pollen to be derived from the intestinal tract of an insect. Although pollen-collecting bees have been reported from Messel [13], the distribution of the pollen all over the putative stomach area argues against their origin from the pollen baskets of hymenopterans.

4. Discussion

For the above reasons, we conclude that the pollen was directly ingested by the bird. The skeletal morphology of Pumiliornis also suggests that it was a, presumably nectarivorous, flower-visitor. The long and slender beak has greatly elongated narial openings [12] (figure 1d), which increase the flexibility of its tip and are also present in hummingbirds, which probe for nectar in deep flowers [14]. The feet of Pumiliornis (figure 1eg) were at least facultatively zygodactyl, that is, the fourth toe could be turned backwards [12]. This morphology is typically found in perching birds, which use their feet to clasp or climb branches and may have facilitated flower-visiting of Pumiliornis. Although pollen size does not allow discrimination of insect and bird pollination [15], the large size of the grains and the fact that some are still clumping (figure 2c) indicate direct ingestion from a plant adapted for animal rather than wind pollination [1].

Flower-visiting evolved independently in three extant avian groups, that is, Trochilidae, Psittaciformes (e.g. lorikeets and hanging parrots) and Passeriformes (various groups, including sunbirds, flowerpeckers, honeycreepers and honeyeaters) [3,4,16]. Most species of these taxa feed predominantly on nectar, but some lorikeets also take the pollen itself. Pollen ingestion has also been reported for hummingbirds, where pollen may amount to 15% of the stomach volume [17]. Occasionally, pollen is also taken by opportunistic flower-visitors [18], but the fact that in one out of only three known skeletons of Pumiliornis pollen is preserved as stomach contents suggests that flower visits were part of the foraging strategy of this bird. The presence of insect remains in the stomach contents is not surprising, as insects today are regularly ingested by nectarivorous birds [17,19].

Sunbirds (Nectariniidae) are the most widely distributed and most species-rich group of Old World avian nectarivores, and the plants visited by these birds alone cover some 450 species in more than 100 families [19]. The tricolpate–tectate pollen ingested by Pumiliornis is clearly derived from eudicotyledonous angiosperms but has previously not been recorded from Messel [10]. We could not find a perfect match among the fossil and extant pollen known to us, but similar pollen occurs in a number of distantly related extant angiosperms, such as Fabaceae (legumes; e.g. Spartium), Lamiaceae (labiates; e.g. Teucrium) and Gesneriaceae (gesneriads; e.g. Ornithoboea). These taxa include ornithophilous species, and Fabaceae as well as tentative records of Gesneriaceae are known as macrofossils from Messel [20].

Middle Eocene avifaunas do not include crown group representatives of any extant avian family-level taxon [9], and P. tessellatus is not closely related to one of the extant nectarivorous birds. Affinities to hummingbirds can be ruled out, because the species lacks diagnostic apomorphies of Apodiformes, such as a greatly abbreviated and stocky humerus. The same is true for closer affinities to Passeriformes, which share numerous derived characteristics that are absent in Pumiliornis (e.g. very narrow tarsometatarsal trochleae and an intermetacarpal process on the carpometacarpus). Among others, absence of a parrot-like beak shows that Pumiliornis is not closely related to any of the crown group Psittaciformes.

The earliest fossil record of modern-type hummingbirds is from the Early Oligocene of Europe [8]. Of the other extant nectarivorous avian taxa, only honeyeaters (Meliphagidae) are represented by fossils, from the Miocene of Australia [21]. All nectarivorous Passeriformes belong to the oscine clade, which has no fossil record before the Late Oligocene [9]. Calibrated molecular phylogenies including a comprehensive sampling of Nectariniidae and Meliphagidae do not exist, but analyses of smaller subsets suggest their diversification after the Eocene [22,23]. Fossils of crown group Psittaciformes are unknown before the Early Miocene [24], and molecular divergence estimates indicate that the flower-visiting parrot taxa likewise evolved after the Eocene [25].

Recognition of a flower-visiting bird in the Middle Eocene of Messel suggests that ornithophilous angiosperm flowers already existed 47 Ma. Because this predates the earliest occurrence of any of the extant nectarivorous avian groups, some angiosperms probably acquired an ornithophilous flower morphology before the origin of their extant avian pollinators. Future time-calibrated molecular phylogenies have to set the timing of the evolution of ornithophilous flowers and their avian pollinators in context, but we consider it likely that some lineages of ornithophilous angiosperms will turn out to be significantly older than their avian flower-visitors.

Our study only provides a minimum estimate for the origin of ornithophily. However, none of the known earlier flowers show adaptations for bird pollination [2], and nectarivory is likewise not suggested by the bill morphology of pre-Eocene birds. Ornithophilous plants exhibit a number of characteristic traits, such as a lack of scent and a red or orange corolla [5], but these are unlikely to be preserved in the fossil record. Bird fossils such as the one described here are thus of key significance for the recognition of bird–flower interactions in past ecosystems.


We thank S. Schaal and E. Brahm for the loan of the fossil, M. Ackermann for its preparation, S. Wedmann for photographs of the insect remains, S. Dimter for assistance with SEM pictures, S. Tränkner for close-up photos and two anonymous reviewers for comments on the manuscript.

  • Received March 13, 2014.
  • Accepted May 2, 2014.


View Abstract