Comparative analysis of vestibular ecomorphology in birds

Abstract

The bony labyrinth of vertebrates houses the semicircular canals. These sense rotational accelerations of the head and play an essential role in gaze stabilisation during locomotion. The sizes and shapes of the semicircular canals have hypothesised relationships to agility and locomotory modes in many groups, including birds, and a burgeoning palaeontological literature seeks to make ecological interpretations from the morphology of the labyrinth in extinct species. Rigorous tests of form-function relationships for the vestibular system are required to support these interpretations. We test the hypothesis that the lengths, streamlines and angles between the semicircular canals are related to body size, wing kinematics and flying style in birds. To do this, we applied geometric morphometrics and multivariate phylogenetic comparative methods to a dataset of 64 three-dimensional reconstructions of the endosseous labyrinth obtained using micro-computed tomography scanning of bird crania. A strong relationship between centroid size of the semicircular canals and body size indicates that larger birds have longer semicircular canals compared with their evolutionary relatives. Wing kinematics related to manoeuvrability (and quantified using the brachial index) explain a small additional portion of the variance in labyrinth size. We also find strong evidence for allometric shape change in the semicircular canals of birds, indicating that major aspects of the shape of the avian labyrinth are determined by spatial constraints. The avian braincase accommodates a large brain, a large eye and large semicircular canals compared with other tetrapods. Negative allometry of these structures means that the restriction of space within the braincase is intense in small birds. This may explain our observation that the angles between planes of the semicircular canals of birds deviate more strongly from orthogonality than those of mammals, and especially from agile, gliding and flying mammals. Furthermore, we find little support for relationships between labyrinth shape and flying style or wing kinematics. Overall, our results suggest that the topological problem of fitting long semicircular canals into a spatially constrained braincase is more important in determining the shape of the avian labyrinth than the specifics of locomotory style or agility. Our results tentatively indicate a link between visual acuity and proportional size of the labyrinth among birds. This suggests that the large labyrinths of birds compared with other tetrapods may result from their generally high visual acuities, and not directly from their ability to fly. The endosseous labyrinths of extinct birds and their close dinosaurian relatives may allow broad inferences about flight or vision, but so far provide few specific insights into detailed aspects of locomotion.

Keywords: birds; endosseous labyrinth; form-function relationships; palaeoecology; phylogenetic signal; semicircular canals.

Figure 1

Rendering of the skull of Podargus strigoides indicating the position of the endosseous semicircular canals. An opaque rendering is shown in (a), and a transparent rendering through which the endosseous canals (minus the vestibule and cochlear canal) are visible is shown in (b). Identity is indicated with colour for the anterior (= rostral) canal (blue), posterior (= caudal) canal (green) and lateral (= horizontal) canal (yellow). Scale bar: 10 mm.

Figure 2

Process of segmenting (a), skeletonising (b) and landmarking (c) the avian labyrinth. The four separate 3D voxel models are shown in grey (entire endosseous labyrinth), blue [anterior (= rostral) semicircular canal], green [posterior (= caudal) semicircular canal] and yellow [lateral (= horizontal) semicircular canal]. Abbreviations: AA, ampulla of the anterior semicircular canal; ASC, anterior semicircular canal; C, cochlear duct; CC, crus communis; LA, ampulla of the lateral canal; LSC, lateral semicircular canal; PA, ampulla of the posterior semicircular canal; PSC, posterior semicircular canal; VE, vestibule. In (C), conventional landmarks at the ends of each semicircular canal are indicated by large yellow spheres and numbered 1–6. Sliding semilandmarks defining the course of each semicircular canal are indicated by small yellow spheres.

Figure 3

Plots of (a) labyrinth centroid size on body mass, with point sizes scaled to brachial index (BI; strong, multivariate relationship; Table 3), (b) labyrinth centroid size on BI (no relationship; Table 3) and (c) residuals of the relationship between labyrinth centroid size and body mass (Table 3, upper panel, model 2, variables log₁₀‐transformed) on BI. Points are coloured according to mutually exclusive subset categories in our multivariate flight categorisation for visualisation purposes only, and are white when none of these categories is present in a species. The decision as to which flight categories should be represented here was informed by the inferred significance of pursuit hunting for labyrinth morphology (see Results). Labyrinth models are shown in lateral view for selected taxa to illustrate the extremes of the relationship of labyrinth size with body mass.

Figure 4

Plots of inter‐canal angles vs. body mass for the angles between (a) the anterior semicircular canal (ASC) and lateral semicircular canal (LSC), (b) the ASC and posterior semicircular canals (PSC), and (c) the LSC and PSC. Points are coloured according to a mutually exclusive subset categories in our multivariate flight categorisation for visualisation purposes only, and are white when none of these categories is present in a species. Labyrinth models are shown in anterior (a), dorsal (b) and lateral (c) views for selected taxa to illustrate the extremes of inter‐canal angles among the sampled taxa.

Figure 5

Landmark configurations corresponding to the mean shape (grey symbols) and deformations along principal component axes PC1–PC3 (black symbols). Deformations correspond to the highest negative (‘min’) and positive (‘max’) score on each PC axis. Deformations are shown in three views corresponding approximately to the orthogonal planes of each of the three semicircular canals. Videos showing these deformations and deformations of other PC axes (PC1–PC12) are available in the Supporting Information.

Figure 6

3D digital models of the endosseous labyrinths of selected species in lateral view, showing morphological changes among species distributed along PC1 (top row), PC2 (middle row) and PC3 (bottom row).

Figure 7

Plots of principal component scores (a) PC1 vs. PC2, (b) PC1 vs. PC3, (c) PC2 vs. PC3, (d) PC2 vs PC4. The proportions of total shape variance explained by each axis are indicated in brackets (Table 5). Points are coloured according to three mutually exclusive categories in our multivariate flight categorisation, and greyed when none of these categories is present in a species. Numbers indicate taxa, as denoted in Table 1. Labyrinth models of selected taxa are shown in lateral view to illustrate extreme values of PC1 and PC2.

Figure 8

Regressions of principal component axes 1–3 (PC1–PC3) on body mass (a, c, e) with point sizes scaled to brachial index (BI), and on BI (b, d, f). PC1 shows a strongly significant relationship with body mass (a) (Table 6), but the other axes shown here show no relationship with either body mass or BI (Tables 6, 7, 8). Points are coloured according to three mutually exclusive categories in our multivariate flight categorisation, and greyed when none of these categories is present in a species.

Figure 9

Interpretation of principal coordinates axis 1 (PCo1) for flying style. The distribution of taxa scored as possessing each flying style (black discs) is shown along the first principal coordinates axis (PCo1; horizontal axis). Flight categories that show clustering towards either end of PCo1 are indicated in black, others are in grey.

Figure 10

Dimension 1 of our non‐allometric two‐block partial least‐squares (2B‐PLS) analysis describes a multivariate association between a composite flying style variable (PCo2) and shape change in the labyrinth described by the principal components axes PC6, PC7 and PC11. (a) The distribution of taxa scored as possessing each flying style (black discs) is shown along the second principal coordinates axis (PCo2; horizontal axis). Flight categories that show clustering towards either end of PCo2 are indicated in black, others are in grey. (b, c) Labyrinth shape deformations (black symbols) implied by change of PC scores along dimension 1 seen in lateral (a) and anterior (b) views, as compared with the mean shape (grey symbols). Key shape differences are increasing sinuosity of the lateral canal, and an increase in the angle between the planes of the anterior and posterior canals at negative shape scores on Dimension 1. (d) Ordination of taxa based on x‐ and y‐scores of dimension 1, showing that two influential sister‐taxon pairs act as influential data points in the analysis. The labyrinth models of these taxa are shown on the plot in lateral view.

Figure 11

Relationship between labyrinth centroid size and body mass (taken from Fig. 3a), with owls and outlier taxon pairs from two‐block partial least‐squares regression (2B‐PLS) indicated. Tyto alba, the barn owl, has a proportionally reduced labyrinth and enlarged cochlea compared with Athene cunicularia, the burrowing owl. 3D virtual models of the labyrinths of A. cunicularia (b, c) and Tyto alba (d, e) are shown in lateral (b), (d) and anterior (c), (e) views to illustrate this.

Figure 12

Spatial constraints on the morphology of the avian semicircular canals, illustrated by 3D scan reconstruction models of (a, b) Scolopax rusticola the woodcock, (c, d) Selasphorus rufus, the rufous hummingbird, and (e, f) Stuthio camelus, the ostrich. (a), (c) and (e) show the endocast (purple) in context of the cranium (grey) in left lateral view. (b), (d) and (f) show closer images in anterolateral view with magnification of the semicircular canals. Scolopax rusticola and Selasphorus rufus show strong evidence of spatial deformation of the semicircular canals due to occupation of space by the brain, Struthio camelus does not. For example, the angle between the planes of the anterior and lateral canals is low in Scolopax rusticola and Selasphorus rufus (indicated), but close to orthogonal in Struthio camelus (indicated).