nt12dreuar3esd

248 | Nature | Vol 579 | 12 March 2020

Article

Oculudentavis follows this pattern and falls where expected in a linear
regression of avian orbit size relative to skull length (Fig.  2 , Extended
Data Fig. 9). The laterally bulging eyes of Oculudentavis, as indicated by
the angled morphology of the jugal bar, probably represent an alterna-
tive strategy for increasing eye size without further increasing the size
of the orbit. Notably, Oculudentavis lacks the forward-facing position
of the eyes of owls and living birds of prey; the eyes are laterally ori-
ented such that binocular vision was absent, or largely limited. Given
the unusual morphologies observed in HPG-15-3, extant analogues for
the visual abilities of Oculudentavis may not exist.
Although miniaturization is often associated with edentulism^9 ,
Oculudentavis possesses an extensive dentition with unusual features.
First, the acrodont to pleurodont geometry of tooth implantation
differs from the thecodont condition in all other dinosaurs. Second,
the tooth row extends ventral to the orbit; this condition is otherwise
only observed among coelurosaur theropods in the Late Cretaceous
ornithurine bird Ichthyornis^12 ,^26. In most coelurosaurs (including Aves),
the maxillary tooth row typically extends as far caudally as the rostral
half of the antorbital fenestra (well cranial to the orbit^27 ), although the
tooth row extends nearly to the jugal–lacrimal contact in some troodon-
tids^28 and in Ichthyornis one tooth is located caudal to the lacrimal and
beneath the orbit. The tooth row is even more caudally extensive in
Oculudentavis, which has four maxillary teeth underlying the orbit.
HPG-15-3 has more teeth than any other known Mesozoic bird, which
highlights both the importance of dentition in the feeding strategies
of Cretaceous avians and the existence of strong positive selection for
teeth despite the overall avian trend towards edentulism^29. Increased
dentition in Cretaceous birds has previously been correlated with pis-
civory^29. However, the papillae preserved on the tongue and ventral
surface of the upper jaw of Oculudentavis (Extended Data Fig. 2a, b)

lack the long, sharp and caudally oriented morphology that is present in

many extant piscivorous birds^14. Although unexpected for its small size,

the large number of teeth—as well as their sharp and carinated morphol-

ogy—suggests that Oculudentavis was a predator, probably feeding on

small arthropods and other invertebrates (which are abundant in amber

inclusions from the Angbamo site). The evolution of island nanism has

been linked to a previously unrecognized ecological guild specialized

in feeding on small invertebrates in Paedophryne frogs (a group that

includes the smallest known vertebrate in the world^30 ), and a similar

diet is possible for Oculudentavis.

Oculudentavis retains the plesiomorphic diapsid condition that is pre-

sent in most nonornithothoracine birds, whereas the infratemporal bar

was independently lost among enantiornithines and ornithuromorphs^27.

Retention of the postorbital bar suggests limited cranial kinesis^19 , consist-

ent with the absence of a distinct craniofacial hinge and with data that sug-

gest avian cranial kinesis is mostly a morphological novelty of neognaths^31.

0

0.2

0.4

0.6

0.4 0.8 1.2 1.6

Geomm

Opt

HPG-15-3

left + right eye

YixianornisArchaeopteryx

Sapeornis

Confuciusornis

Fig. 3 | A star plot illustrating the functional morphospace of scleral-ring
and orbit morphology of modern saurians in the context of their diel
activity pattern. The x axis (labelled ‘geomm’) represents the log 10 -
transformed geometric mean of orbit length and the external and internal
scleral-ring diameters, and the y axis (labelled ‘opt’) is formed by the optical
ratio. Triangles indicate squamates (n = 60 species); circles indicate birds
(n = 104 species). Ellipses represent 95% confidence areas for cathemeral
(green), diurnal (orange) and nocturnal (purple) groups, drawn around the
respective average x and y values for each group. Density plots illustrate the
distribution of data along the x and y axes, delineated by groups of diel activity
pattern. The retrodeformed eyes of HPG-15-3 are most similar in shape and size
to the eyes of diurnal squamates. The black oval around the data points
representing the left and right eyes of HPG-15-3 represent the 95% confidence
ellipse, indicating the range of values estimated by our sensitivity analysis. The
average of the geometric mean of all eye variables and the optical ratio for the
retrodeformed left and right eyes were calculated and used to generate a
distribution of possible combinations spanning ±30% of the left- and right-eye
averages.

Pygostylia

Av

es

Ornithurae

Ornithothoraces

Ornithur

omorpha

Enantior

nithes

Dromaeosauridae

Archaeopteryx

Oculudentavis

Jeholornis

Confuciusornithiformes

Sapeornis

Concornis

Eocathayornis

Eoalulavis

Cathayornis

Halimornis

Enantiophoenix

Shenqiornis

Neuquenornis

Eoenantiornis

Elsornis

Pengornithidae

Gobipteryx

Longipterygiformes

Protopteryx

Iberomesornis

Archaeorhynchus

Patagopteryx

Longicrusavis

Yanornis

Gansus

Apsaravis

Ichthyornis

Anas

Gallus

Hesperornis

Fig. 4 | Simplif ied results of the strict consensus of 2,044 trees depicting the

phylogenetic relationships of O. khaungraae relative to other known

Mesozoic birds. The analysis (consisting of 36 taxa scored across

257 characters) was run in the TNT program using implied weighting (k = 16).

Consistency index = 0.47; retention index = 0.645. In the analysis using a priori

weights, Oculudentavis was resolved as an enantiornithine in a small percentage

of the most-parsimonious trees. For most species for which skull material is

known, the skull is reconstructed to scale relative to O. khaungraae. Scale bar,

2 cm. Names of species for which no skull material is known are written in grey.