The endocast of the Night Parrot (Pezoporus occidentalis) reveals insights into its sensory ecology and the evolution of nocturnality in birds

The Night Parrot (Pezoporus occidentalis) is a rare, nocturnal parrot species that has largely escaped scientific investigation due to its behaviour and habitat preferences. Recent field studies have revealed some insights into Night Parrot behaviour, but nothing is known of its sensory abilities. Here, we used μCT scans of an intact Night Parrot specimen to determine if its visual system shares similarities with other nocturnal species. The endocast of the Night Parrot revealed relatively small optic lobes and optic foramina, especially compared with closely related grass parakeets, but no apparent differences in orbit dimensions. Our data suggests that the Night Parrot likely has lower visual acuity than most other parrots, including its congener, the Eastern Ground Parrot (P. wallicus). We propose that the visual system of the Night Parrot might represent a compromise between the need to see under low light conditions and the visual acuity required to detect predators, forage, and fly. Based on the endocast and optic foramen measurements, the Night Parrot fits into a common pattern of decreased retinal input to the optic lobes in birds that should be explored more thoroughly in extant and extinct species.

of the brain become smaller [13][14][15] . In the only other nocturnal parrot, the Kakapo (Strigops habroptilus), eye size and shape do not differ from diurnal parrots, but the optic tectum, the primary target of retinal ganglion cells, is greatly reduced in size 16 . The shrinkage of the Kakapo optic tectum coincides with a decrease in size of the optic foramen, which houses the optic nerve, as well as fewer retinal ganglion cells 16 . These anatomical changes result in the Kakapo possessing a visual system with a greater ability to capture photons under low light (higher sensitivity), but a relatively poor ability to discriminate among visual stimuli (lower acuity) compared to diurnal parrots 17 . Assessing whether the Night Parrot shares a similar visual system to the Kakapo would yield new insights into Night Parrot behaviour, including a better understanding of how they perceive their habitat.
Detailed study of the Night Parrot eye and brain anatomy is not possible due to the extremely limited number of living individuals and lack of fluid preserved museum specimens. However, much information regarding the Night Parrot's visual system can be gleaned from the skull. For example, the optic foramen is well defined in parrots 18 and reflects optic nerve size 16,18 and orbital measurements can approximate eye size 19 . Digital endocasts (three dimensional reconstructions of the brain based on μCT scanning of skulls) have also been useful in assessing the sensory ecology of extinct birds [20][21][22][23] . In particular, the surface area of the optic lobes reflects the volume of the underlying optic tectum 24 , the midbrain region that receives the majority of retinal projections in birds 25 . The optic tectum is also the region that undergoes the greatest reduction in size in other nocturnal birds 15,21,26 , including the Kakapo 16 . Thus, the relative size of the optic foramen, orbits, and optic lobes of the Night Parrot could provide insights into its visual abilities.
Here, we used μCT scans of the only Night Parrot skull known to be intact 3 (Fig. 1a) to quantify the dimensions of the orbits, optic foramen and optic lobes in comparison with other parrot species. Specifically, we predicted that the Night Parrot would have reduced optic lobes and smaller optic foramina, similar to the Kakapo 16 . However, we expected a reduction of lesser magnitude than what was found in the Kakapo because Night Parrots fly considerable distances between roosting and feeding areas 6 , probably making them more dependent on vision 27 .

Results
The μCT scans of the Night Parrot specimen (Fig. 1a) revealed some fragmentary material on the outside of the skull, but the braincase and orbits were entirely intact, allowing us to complete measurements of both the orbits and the endocast (Fig. 1b). Overall, the endocast had a volume of 2,478.07 mm 2 , which relative to body mass is typical of other small parrots (Fig. 1c) and similar to that of the Eastern Ground Parrot (Pezoporus wallicus). In terms of morphology, however, the optic lobes of the Night Parrot (Fig. 2a) appeared to be unusually small (Fig. 2b,c), especially compared with its congener and closest relative within the sample, the Eastern Ground Parrot (Fig. 2d-f), and the Bourke's Parrot (Neopsephotus bourkii, Fig. 2g-i), a closely related grass parakeet 28 that lives in similar habitat to the Night Parrot and is often active at dusk 29,30 .
Compared with total endocast surface area, the Night Parrot had the smallest optic lobes of any of the species we examined (Fig. 3a,b). Although the Purple-crowned Lorikeet (Glosspsitta porphyrocephala) also has small optic lobes, when expressed as a percentage of total endocast surface area, the Night Parrot had the lowest value (Fig. 3b). More importantly, the relative size of the Night Parrot's optic lobes contrasts greatly with that of its relatives, the grass parakeets (Neophema, Neopsephotus), and its congener the Eastern Ground Parrot, which had the largest optic lobes relative to endocast surface area (Fig. 3a,b). As shown in Fig. 3c, the optic lobes of the Night Parrot falls outside of the 95% interval of our phylogenetically informed posterior probability distributions. In fact, it exceeds the 99% interval, indicating that the surface area of the Night Parrot's optic lobes is far below that predicted by its total endocast surface area.
The Night Parrot also had the smallest optic foramen relative to endocast volume (Fig. 3d). In fact, the optic foramen of the Night Parrot was similar in absolute size to that of the Budgerigar (Melopsittacus undulatus), a species with an endocast volume 2/3 that of the Night Parrot. At the other end of the spectrum, the Eastern Ground Parrot had an optic foramen that was almost 2x that of the Night Parrot even though they share similar endocast volumes (Fig. 3d). The relatively small size of the Night Parrot's optic foramen is also supported by the posterior probability distribution (Fig. 3e).
Unlike the optic lobes and optic foramen, the Night Parrot did not differ in relative orbit area or depth from other parrots (Fig. 3f-i). Although the Ground Parrot did appear to have enlarged orbit area, relative to endocast volume (Fig. 3g), it did not fall outside of the 95% credible interval (Fig. 3h).

Discussion
Overall, the Night Parrot has undergone a decrease in relative optic lobe size, at least compared with the closely related grass parakeets, and a decrease in relative optic foramen diameter compared with all other parrots examined. This indicates that the Night Parrot has evolved significant changes in its visual system anatomy, likely in response to its nocturnal activity. Although our conclusions are based on a single Night Parrot specimen, this is one of the rarest species in ornithological collections worldwide 1,31 . No osteological specimens appear to exist 1,3,31 and the preparation of study skins involves damage to or removal of the skull 32 . The 'mummified' specimen 3 is therefore the only Night Parrot specimen known to have an intact skull. Further, for species in which we measured three specimens, intraspecific coefficients of variation were 0.06-0.11 across the optic lobe and optic foramen measurements, suggesting that intraspecific variation is relatively low. If the Night Parrot's optic lobe area and optic foramen diameter are underestimated by 11%, the observed values would still fall outside of the 95% credibility intervals in Fig. 3c,e. Thus, based upon the available data, we are reasonably confident that the Night Parrot has undergone reductions in the optic lobes and optic foramen, especially in comparison to the Eastern Ground Parrot and other grass parakeets. Despite the reductions in optic lobes and foramina, the Night Parrot did not differ in orbit size from the other species examined. Eye shape, corneal diameter and retinal morphology are often better predictors of low light vision than eye size 10,18,19 . For example, the nocturnal Kakapo does not differ from other parrots in eye size or shape, but does have a higher density of photoreceptors and fewer retinal ganglion cells 16 . Currently, it is not possible to estimate photoreceptor density in the Night Parrot because a fluid preserved specimen does not exist and capturing one to preserve appropriately is not a viable option. However, the optic nerve is comprised primarily of retinal ganglion cell axons and the size of the optic foramen closely approximates that of the optic nerve in most birds, including parrots 16,18 . The relatively small optic foramina of the Night Parrot therefore reflects smaller optic nerves and, by extension, fewer retinal ganglion cells in the retina. If this assumption is correct, the Night Parrot likely has lower visual acuity than the other parrot species examined because the spatial resolving power of the eye is a product of retinal ganglion cell density and eye size 33 . www.nature.com/scientificreports www.nature.com/scientificreports/ Corroborating support for fewer retinal ganglion cells in the Night Parrot is provided by the reduced optic lobes. The optic tectum receives the majority of the retinal efferents in most birds 25 and species with fewer retinal ganglion cells also have relatively small optic tectum volumes 16,26,34 . Although the optic lobes house more than just the optic tectum, there is a strong correlation between optic tectum volume and the surface of the optic lobes 24 . Thus, the relatively small optic lobes of the Night Parrot likely reflect a decrease in optic tectum size and go hand in hand with smaller optic foramina.
A potential consequence of smaller optic foramina and optic lobes and fewer retinal ganglion cells is a decrease in visual acuity. Lower visual acuity is typically a consequence of living in scotopic (i.e., low light) environments 35 . In order to see effectively under scotopic conditions, the eye needs to enhance its sensitivity and this comes at the expense of visual acuity 35 . For example, in owls, and to a lesser extent Kakapo, a high population of photoreceptors converge on a lower number of retinal ganglion cells to improve sensitivity at the expense of acuity 16,36,37 . Based on our data, we suggest that the visual acuity of the Night Parrot is lower than that of other parrots, especially the closely related Eastern Ground Parrot and other grass parakeets. Having less acute vision is unlikely to be a hindrance to Night Parrots because they prefer open habitats with few, if any trees 1,6 , and the risk of flying into natural obstacles is low. Lower visual acuity and flying at night could, however, increase the risk of mortality arising from anthropogenic obstacles, such as fences, that are immobile and low contrast 38 . This is likely one of the primary reasons that owls and other nocturnal birds frequently become entangled on barbed-wire fences [39][40][41] . A survey of bird casualties due to barbed-wire fencing in Diamantina National Park, where Night www.nature.com/scientificreports www.nature.com/scientificreports/ Parrots also occur 42 , revealed that at least half of the species entangled are nocturnal or active at dusk 43 . More importantly, a decapitated Night Parrot was also found below a stretch of barbed-wire fencing on the park 4 and the species most commonly killed as a result of fence strike was the crepuscularly active Bourke's Parrot 43 . If we correct in our conclusion that the Night Parrot has lower visual acuity than diurnal parrots, barbed-wire fences could pose a significant hazard to the Night Parrot throughout their range 1 .
Although our interpretation of the Night Parrot endocast is somewhat speculative, the reduction of the optic foramina and lobes corroborates a more generalized relationship between brain morphology and nocturnality in birds. Overall, nocturnal birds have significantly smaller optic foramina than diurnal birds 18 . Many nocturnal birds also have relatively small optic tectum volumes 15,16,26,44 . Similar reductions in the optic lobes and foramina are also reported in extinct birds, such as the elephantbirds (Aepyornis spp.) 21 and Hawaiian 'mole-duck' (Talpanas lippa) 45 , both of which were interpreted as evidence of nocturnal behaviour. The Night Parrot adds to this generalized pattern of reduced optic foramina/nerves and optic lobes/tectum in nocturnal birds. We therefore emphasize that quantitative analyses of endocasts and cranial nerves may have significant potential in determining the activity pattern and sensory abilities of extinct and critically endangered bird species in ways that have thus far been largely overlooked.

Materials and Methods ethics Statement.
No animals were collected for use in this study. All measurements were made from specimens housed at the Queensland Museum (see Table 1).

Specimens.
To examine the anatomy of the endocast and orbits of the Night Parrot, we μCT scanned the specimen that was found in southwest Queensland in 1990 3 (Fig. 1a), the only specimen known to have an intact skull. For comparison, we scanned a total of 22 specimens representing 17 other parrot species. Within the monophyletic clade Pezoporini 46 , to which the Night Parrot belongs 28 , we examined several grass parakeets (Neophema spp.), Bourke's Parrot and the congeneric Eastern Ground Parrot. Bourke's Parrots are frequently active after sunset 29,30 and differ in vision and photoreceptor densities from exclusively diurnal parrot species 47,48 . The Ground Parrot shares similar terrestrial habits to the Night Parrot, but is diurnal and prefers coastal heathlands and sedgelands rather than arid interior regions 30,49,50 . The other 15 species examined were selected based on availability, size and phylogenetic distribution so that we had a sufficient range of sizes to calculate allometric relationships and had species representing clades other than Pezoporini within the Psittaciformes 46 .
The skulls were scanned using high-resolution X-ray computed tomographic scans (μCT) from a Siemens Inveon PET/CT scanner at the Centre for Advanced Imaging, University of Queensland. The resulting DICOM files were imported into Mimics (v18.0, Materialise NV) and skull 'masks' were produced through thresholding (Fig. 4). Segmenting the endocast was done slice-by-slice ensuring that foramina and fenestrae were segmented www.nature.com/scientificreports www.nature.com/scientificreports/ consistently across specimens 51 . A complete list of the species, specimen numbers and data are provided in Table 1. All original DICOM scans with associated acquisition parameters, as well as the surface files of cranial reconstructions and endocasts, can be found on MorphoSource (Project P453).
Measurements. 3D reconstructions of the skulls were used for measuring the surface area of the orbit, optic foramen and foramen magnum as well as the orbit depth within Mimics. All of the quantitative measurements are provided in Table 1. Orbit diameter was assessed through use of the ellipse tool. The ellipse was placed to touch the left, right, upper, and lower extremes of the orbital rim, and the surface area of the resulting (mostly circular) ellipse was determined. Similarly, the ellipse tool was used to outline the optic foramen and determine its area. Orbit depth was measured from the dorsal rim of the optic foramen to the center of the plane described by the ellipse fitted across the orbit, the same ellipse that was used to measure orbit circumference (see Figure S1 in supplementary information).
Acquiring the surface area of the optic lobes required a stepwise approach. The trigeminal nerve and blood vessel located on the surface of the optic lobe were digitally removed to avoid overestimates of optic lobe area. After this, the optic lobes were digitally dissected from the rest of the endocast and data was recorded for each side and then averaged. As the total surface area of the lobes included two sides where they had been cut from the rest of the endocast, the ellipse tool was used to measure these areas, so they could be subtracted from this total. The original left and right measurements are provided in the supplementary information (Table S1). For species represented by more than one specimen, data were averaged for analysis (Table 1).
Mesh volumes, using the "optimal resolution" setting in Mimics, were used to facilitate the dissection of the optic lobes ( Figure S2 in supplementary information). The conversion of mesh volumes from voxel volumes produces slight discrepancies between the voxels and the mesh outlines, potentially leading to very small deviations of voxel-based vs. mesh-based volume estimates ( Figure S2). Relative to the volumes measured, this effect is very small, identical across dissections, and does not result in distortion of the overall shape. It was therefore deemed negligible as a source of error.

Statistical analysis.
We first plotted endocast volume against body mass of the Night Parrot and a large dataset of endocranial volumes of 180 other parrot species 52 . Body mass for the Night Parrot is the average of two specimens that were captured for a GPS tracking study 6 and for other species from 52 . We then used a phylogeny informed statistical approach to test whether the Night Parrot deviated from allometric relationships for optic lobe surface area, optic foramen area and orbit measurements. The phylogeny was constructed from published papers 28,53 using Mesquite 54 (Fig. 4). Because the phylogeny was constructed from different sources, we set all branch lengths to one. To then evaluate whether the Night Parrot differed in the relative size of the optic lobes, optic foramen, and orbits, we generated posterior probability distributions of the expected values using a Bayesian Markov Chain Monte Carlo approach across the phylogeny, following the procedures outlined in Nunn and Zhu 55 and performed in R 56 . We ran 200,100 iterations with a burnin rate of 100 and a thin rate of 100 with endocast volume as the predictor variable to generate posterior probability distributions of 2,000 values of the predicted sizes of the optic lobes, optic foramen, and orbit dimensions of the Night Parrot.