Variation in air sac morphology and postcranial skeletal pneumatization patterns in the African grey parrot

Abstract

The anatomy of the avian lower respiratory system includes a complex interaction between air‐filled pulmonary tissues, pulmonary air sacs, and much of the postcranial skeleton. Hypotheses related to the function and phylogenetic provenance of these respiratory structures have been posed based on extensive interspecific descriptions for an array of taxa. By contrast, intraspecific descriptions of anatomical variation for these features are much more limited, particularly for skeletal pneumatization, and are essential to establish a baseline for evaluating interspecific variation. To address this issue, we collected micro‐computed tomography (μCT) scans of live and deceased African grey parrots (Psittacus erithacus) to assess variation in the arrangement of the lungs, the air sacs, and their respective invasion of the postcranial skeleton via pneumatic foramina. Analysis reveals that the two pairs of caudalmost air sacs vary in size and arrangement, often exhibiting an asymmetric morphology. Further, locations of the pneumatic foramina are more variable for midline, non‐costal skeletal elements when compared to other pneumatized bones. These findings indicate a need to better understand contributing factors to variation in avian postcranial respiratory anatomy that can inform future intraspecific and interspecific comparisons.

The avian lung‐air‐sac system invades much of the skeleton with epithelium‐lined extensions of their air‐filled spaces through a process called pneumatization. This study uses micro‐computed tomography scans of the African grey parrot to identify and describe: (1) intraspecific variations in air sac arrangement and (2) the location of the bony openings (called pneumatic foramina) that connect the lungs to these spaces.

1. INTRODUCTION

The avian gas‐exchanging lungs are immobilized (Duncker, 1971; Maina, 2007b), nearly volume‐constant (Jones et al., 1985; Makanya et al., 2020), and are directly attached to the dorsal vertebrae and ribs (McLelland, 1989; Schachner et al., 2009; Schachner et al., 2021). Prior to hatching, birds develop saccular extensions from their bronchial airways that form and fuse to become a series of compliant, ventilatory air sacs (Locy & Larsell, 1916a; Locy & Larsell, 1916b). These air sacs fill much of the thoracocoelomic cavity (Duncker, 1979; King & McLelland, 1984) and include diverticular extensions (“pulmonary and pneumatic diverticula,” O'Connor, 2006) that variably surround and appose viscera and bone (Baumel et al., 1993; King, 1966; McLelland, 1989). Further invasion by these pulmonary tissues into bone occurs during an approximate six‐month period post‐hatching (Bremer, 1940b; Hogg, 1984a; Schepelmann, 1990) via pneumatic foramina that provide passages through cortical bone to connect avian respiratory tissues with epithelia‐lined, air‐filled spaces in the adult skeleton (O'Connor, 2006; Witmer, 1990). The extent to which respiratory tissues invade the individual skeletal elements is highly variable, and has been described for numerous adult taxa as including the humeri, coracoids, sternum, pelvis, ribs, and vertebrae, extending as far cranially as the second cervical vertebra (Hogg, 1984b; King, 1957; King, 1966; O'Connor, 2004). The intricate interface of respiratory tissues, viscera, muscle, and bone is consequently complex, and the functional roles of this interface are still poorly understood (Atterholt & Wedel, 2023; Hogg, 1984a; Maina, 2007a; Maina, 2017; Maina et al., 2021; O'Connor, 2004; O'Connor, 2006; Schachner et al., 2021; Wedel, 2007).

Comparisons of pulmonary anatomy reconstructions for extinct and living taxa have long been pursued in the context of extant phylogenetic bracketing (Witmer, 1995) and, relatedly, as a means to establish the diverse functional roles played by these tissues (Atterholt & Wedel, 2023; Aureliano et al., 2022; Butler et al., 2012; Gutherz & O'Connor, 2021; Gutzwiller et al., 2013; King, 1966; McLelland, 1989; O'Connor, 2006; Wedel, 2006). Air sacs and bronchi, for example, have both been analyzed and modeled (Banzett et al., 1987; Brackenbury, 1979; Brown et al., 1995; Harvey & Ben‐Tal, 2016; Hazelhoff, 1951; Maina et al., 2009; Nguyen et al., 2021; Wang et al., 1988) to determine their role in sustaining observed unidirectional airflow patterns in birds (Banzett et al., 1991; Fedde, 1998; Powell et al., 1981; Scheid, 1979; Scheid et al., 1972; Wang et al., 1992). Unidirectional airflow, however, has also been observed in other vertebrate taxa that lack air sacs (Cieri et al., 2014; Cieri & Farmer, 2019; Farmer, 2015; Schachner et al., 2013; Schachner et al., 2014), and unidirectional airflow patterns are sustained in birds during systematic in vivo physiologic air sac occlusion (Brackenbury et al., 1989; Brackenbury & Amaku, 1990a; Brackenbury & Amaku, 1990b). These findings suggest that avian air sacs are not integral for unidirectional airflow and that their morphology may be influenced by undetermined ecological, ontogenetic, and/or physiologic constraints. Air sacs vary between taxa as well, and have been hypothesized and/or demonstrated to serve several non‐respiratory functions including vocalization (Lockner & Murrish, 1975; Plummer & Goller, 2008), visual display (Akester et al., 1973), thermoregulation (Menuam & Richards, 1975; Schmidt‐Nielsen et al., 1969; Soum, 1896; Wedel, 2003), maintenance of normocapnia (Maina & Nathaniel, 2001), protection from trauma (Daoust et al., 2008; Richardson, 1939), and modification of flight muscle mechanics in soaring taxa (Schachner et al., 2024). Pneumatization patterns, on the other hand, have been suggested to correlate with body size (Benson et al., 2012; Moore, 2021) and/or reduce the metabolic costs of locomotion (Burton et al., 2023; Farmer, 2006) for taxa that rely on cursorial locomotion (Gutherz & O'Connor, 2021), static soaring flight (Gutherz & O'Connor, 2022; O'Connor, 2009), or aquatic pursuit diving (O'Connor, 2004; O'Connor, 2009; Smith, 2012). Rather than lightening the overall skeleton (Prange et al., 1979), postcranial skeletal pneumaticity redistributes mass and adjusts overall mass–volume relationships (Dumont, 2010). The extent of this pneumaticity has been reported to range from complete absence, as in several anseriform taxa (O'Connor, 2004), to pneumatization of nearly the entire skeleton, as in specific members of Anhimidae (Demay, 1940). Identification of similar pneumatic features in the fossil record has additionally informed functional hypotheses related to pterosaurs (Claessens et al., 2009; Martin & Palmer, 2014), sauropods (Wedel, 2003; Yates et al., 2012), non‐avian theropods (O'Connor, 2006; O'Connor & Claessens, 2005), and other ancestral vertebrate taxa (Farmer, 2017; Lambertz et al., 2018).

Interspecific comparisons involving small sample sizes often rely on an implicit assumption articulated by Müller (1908) that pneumatic foramina have intraspecifically consistent arrangements and locations. In contrast, intraspecific anatomical variation has been often described for other cardiopulmonary and vertebral features including air sac arrangement (e.g., Casteleyn et al., 2018; Daoust et al., 2008; Duncker, 1971; Gier, 1952; Hamlet & Fisher, 1967; Homberger, 2017; King, 1966), overall extent of avian skeletal pneumatization (Biur & Thapliyal, 1972; Burton et al., 2023; Crisp, 1857; Hogg, 1980; Hogg, 1990; King, 1966; Leberman, 1970; McNeil & Jean, 1972; Schepelmann, 1990; Witmer, 1990), nutrient foramina location in extant vertebrates (De Buffrénil et al., 2008; Smuts, 1975), pneumatic foramina locations in fossils of non‐avian dinosaurs (Hatcher, 1901; O'Connor, 2007; Taylor & Wedel, 2021; Wedel & Taylor, 2013), and pneumatic foramina locations in extant avian taxa (Apostolaki et al., 2015). Histological descriptions of air sac pneumatization in the cranium (Bremer, 1940a) and humerus (Bremer, 1940b) of chicks (Gallus gallus domesticus) describe variability in the growth and appearance of epithelium‐lined tubular projections that invade perichondrium and maturing periosteal bone, noting of the air sacs that “apposition [to bone] is not sufficient to ensure entrance.” Bremer further notes the formation and occasional degeneration of these projections and the accompanying mesenchyme, osteoblasts, and veins.

Conclusions from these collected reports on gross respiratory anatomy align with Locy and Larsell's (1916b) remarks when describing avian lung embryology that, “individual variation is very common,” and prompt the question: How much are the observed variations in avian pulmonary anatomy a consequence of variations in the timing of tissue growth and development? Skeletal pneumatization in vertebrates has been proposed to be opportunistic (Witmer, 1997), dictated by the local conditions of the invading respiratory element, adjacent blood supply (Anderson‐Berry et al., 2005; Hönig et al., 2002; Ikarashi et al., 1996), and biomechanically constrained architecture of the invaded bone (Farke, 2010; Moore, 2021; Rae & Koppe, 2008). Genetic variation has also been implicated in pneumatic variation (Ito et al., 2015). Since pneumatic invasion of the skeleton requires the presence of bone‐resorbing osteoclasts and accompanying nutrient blood vessels (Bremer, 1940b; O'Connor, 2006), variation in air sac morphology and vascular supply (Berger, 1956) would a priori be expected to generate variations in the extent of adult bone pneumatization and pneumatic foramina locations.

Establishing a baseline for the extent of intraspecific variation is confounded (1) by a relative lack of intraspecific comparisons providing detailed analyses of features like air sac morphology (Atterholt & Wedel, 2023), bronchial tree arrangement (Fischer, 1905; Lawson et al., 2021; Maina & Nathaniel, 2001; Schachner et al., 2021), or postcranial skeletal pneumatization (Campana, 1875; Gutzwiller et al., 2013; Hogg, 1984a; King, 1957; King & Kelly, 1956) and (2) by the methodological heterogeneity of existing interspecific comparisons. King (1966) and McLelland (1989) both remarked on the unreliability of the various studies that—in addition to differing terminology standards, visualization techniques (e.g., dissection vs. latex endocasts), and criteria for pulmonary feature identification—often rely on analysis of fragile pulmonary tissues, friable pulmonary endocasts, and ex vivo bones. Pneumatization, for example, has been described as being exclusively provided post‐hatching by pneumatic diverticula from the air sacs (Hogg, 1984a; King, 1957; King & Kelly, 1956) or, contrastingly, as also being provided by “pulmonary diverticula” (O'Connor, 2006) arising from the lung airways (Atterholt & Wedel, 2023; Duncker, 2004; Hunter, 1774; Müller, 1908; O'Connor, 2006; Wedel, 2009). Similar ambiguities confound distinctions of pneumatic bone from non‐pneumatic bone through gross features (Apostolaki et al., 2015; Gutzwiller et al., 2013; Müller, 1908; O'Connor, 2006). The use of trabecular bone density as a criterion (Bellairs & Jenkin, 1960; Bremer, 1940b), for example, has been shown to be unreliable for determining the pneumaticity of the third thoracic vertebra in two similarly sized anatid taxa (Fajardo et al., 2007). Relatedly, recent work with fresh, frozen avian humeri has indicated that previous assessments may have broadly overestimated the proportions of bone volumes that are occupied by air (Burton et al., 2023). Conflicting reports of pneumatization may also involve variation in which bony correlates are included as a criterion, such as concavities apposed to respiratory air sacs (“noncommunicating fossae,” O'Connor, 2006) that do not penetrate the cortex (Lambertz et al., 2018; Mayr, 2021).

To better inform existing and future hypotheses of function, homology, and phylogeny, this study examined intraspecific variability in the adult morphology of the air sacs, pneumatization patterns, and pneumatic foramen location in the African grey parrot (Psittacus erithacus). Micro‐computed tomography (μCT) scans of P. erithacus were analyzed to address the following hypotheses: (1) reduced functional constraints on the ventilatory air sacs relative to the bronchial airways (Lawson et al., 2021) will generate greater intraspecific variability in air sac morphology that will (2) influence the intraspecific variability of skeletal pneumatization extent and pneumatic foramen location. Descriptive analyses from these data establish a robust starting point for further in situ analyses of interactions between vertebrate postcranial respiratory and skeletal elements. Additionally, analysis and three‐dimensional (3D) reconstructions of these data provide clinically relevant information for veterinary imaging (McMillan, 1994; Silverman & Tell, 2010) and surgical intervention (Echols, 2018; Rubin et al., 2016; Speer, 2018; Taylor, 2016) of an endangered taxon (BirdLife, 2018).

2. MATERIALS AND METHODS

2.1. Materials

The μCT scans of African grey parrots (P. erithacus) used in this study included both deceased companion birds (previously described in Lawson et al., 2021) that were donated to the Grey Parrot Anatomy Project (n = 7) and live, adult birds from the Tracy Aviary in Salt Lake City, Utah (n = 2). Owners of the deceased parrots signed a release form donating the bird's postmortem to the Grey Parrot Anatomy Project, acknowledging their use for research. Although specimens exhibited signs of skeletal disease (e.g., osteopenia) that commonly affect P. erithacus (Ritchie et al., 1994), the etiology of included birds, living or dead, did not include respiratory disease, and each was renamed to preserve anonymity (Table 1). The deceased birds were intubated via the glottis, inflated to maximum inspiratory capacity, and scanned at the University of Utah (see Lawson et al., 2021 for further description). Live adult birds were sedated with intramuscular injection of butorphanol (2 mg/kg) and anesthetized via a facemask with isoflurane/O₂ (O₂ 2 L/min). Both in vivo scans were imaged in the Fall of 2018 on an Epica Vimago veterinary μCT scanner (slice thickness = 200 μm) by employees of the Parrish Creek Veterinary Hospital and Diagnostic Center in Centerville, Utah. No approval from the Institutional Animal Care and Use Committee was needed for the live birds included in this study because they were imaged for clinical purposes unrelated to this research, and the scans were later donated to this project. All birds were scanned in dorsal recumbency, except for Plato who was scanned in ventral recumbency. All imaging data are available via MorphoSource in DICOM or TIFF format: https://www.morphosource.org/projects/000554139.

TABLE 1.

P. erithacus specimens used for the air sac analysis and pneumatic foramen identification.

Name	Status	Scan	Position	Slice (μm)
Aristophanes	Deceased	Full	Dorsal	97.193
Aristotle	Deceased	Full	Dorsal	94.185
Homer	Deceased	Full	Dorsal	94.383
Plato	Deceased	Body	Ventral	97.230
Pythagoras	Deceased	Full	Dorsal	94.185
Socrates	Deceased	Full	Dorsal	94.185
Zeno	Deceased	Full	Dorsal	94.185
Diogenes	Live	Full	Dorsal	200.000
Hesiod	Live	Full	Dorsal	200.000

Note: Blue highlighting indicates specimens that were selected for pneumatic foramen identification based on slice thickness and lack of scanning artifacts.

2.2. Methods

3D models of the lower respiratory system, axial skeleton, and skeletal pneumatization patterns were generated using Avizo Lite 2020 (Thermo Fisher Scientific). Air sacs and more radiopaque portions of the skeleton were segmented via semi‐automated grayscale thresholding by adaptively adjusting the thresholding tool in every slice to select the boundaries between air, bone, and/or soft tissues. When the contrast was insufficient, boundaries of the air sacs, skeleton, and diverticula were defined through manual segmentation with a Wacom Intuos Pro pen tablet following the methods outlined in Lawson et al. (2021), Schachner et al. (2021), and Schachner et al. (2023). User‐driven manual and threshold‐based techniques were semi‐automated by variably interpolating 1–10 intermediate slices. Variations in boundary choice produced by the combination of semi‐automated and user‐driven techniques produce a “terracing” artifact in the 3D surface models. Following initial work in the axial plane of the original scans, Avizo's two‐dimensional (2D) reconstructions of sagittal and horizontal (dorsal) planes were examined to inform iterative revisions of the segmentation. The boundaries of the paired cervical and unpaired interclavicular sacs were indistinguishable and so were segmented to reconstruct as one continuous 3D surface model.

While all the scans had sufficient resolution to locate and compare the arrangement and inflation of the voluminous air sacs in deceased and live birds (Table 1), the in vivo scan resolution was insufficient to identify pneumatic foramina. Analysis of pneumaticity was therefore restricted to the deceased birds that were further selected (i.e., Aristophanes, Aristotle, Homer, Plato, and Zeno) for showing no signs of pulmonary decay and for exhibiting the least image artifacts (e.g., motion blur, streaking due to beam hardening). The location of visible pneumatic foramina in these scans was identified for the entire postcranial skeleton using the multi‐planar reconstruction (MPR) module available in OsiriX MD (Pixmeo SARL). Pneumatic foramina were identified by authors ABL and AM and assessed for connection to either specific air sacs or bronchial airways. The criterion that defined a pneumatic foramen was an unambiguous break in the cortical bone that could be traced through multiple planes connecting a radiolucent (air‐filled) pulmonary element to a radiolucent (pneumatized) space within a skeletal element. Ambiguous and/or radiopaque breaks were recorded when observed, but not considered for final analysis.

Terms used to describe airways (e.g., primary bronchi, ventrobronchi, dorsobronchi, laterobronchi, and parabronchi; Duncker, 1971) and air sacs (i.e., cervical, cranial thoracic, caudal thoracic, abdominal, and interclavicular sacs; King, 1966) are based on precedents set by Schachner et al. (2021). Terms defined by Baumel et al. (1993) were used for particular bones and bony landmarks. Excepting the supramedullary diverticulum, McLelland (1989) provides the precedent for the names and criterion descriptions of air sac diverticula. The term “supramedullary” is based on the precedent established by Müller (1908) and maintained by O'Connor (2006), although a survey has recently suggested the use of “paramedullary” to account for variations in many taxa that extend around and beyond the vertebral canal (Atterholt & Wedel, 2023). Descriptions of the air sacs were gathered by examining the μCT scans of deceased and live birds in conjunction with the 3D surface models generated from the scans.

3. RESULTS

3D reconstruction and analysis of the μCT image stacks reveal variability in air sac arrangement, skeletal pneumatization, and pneumatic foramen location in the different specimens of P. erithacus.

3.1. Air sac morphology

Excepting Aristophanes (see below), five of the six specimens exhibit a typical arrangement of nine air sacs extending from the airways of the lung (Figure 1): seven air sacs extending into the subpulmonary cavity, and two into the intestinal part of the coelom. The subpulmonary cavity contains an unpaired interclavicular sac and three pairs of sacs: cervical, cranial thoracic, and caudal thoracic. Although regions of the interclavicular and cervical sacs have indeterminable boundaries with one another, they are nevertheless identifiable using μCT by tracking their relatively unambiguous connections to secondary bronchi and diverticula. The remaining six air sacs are generated as independent 3D models separated by distinct soft tissue planes in the scans that represent the thin, apposing epithelial walls of the air sacs. Abdominal air sacs are assumedly sequestered to the abdominal portion of the coelom by a well‐described, non‐muscular membrane called the oblique septum (Duncker, 1979), although scan resolutions were not sufficient to distinguish this soft tissue structure from the membranous surfaces of the air sac. All the sacs have connections with specific bronchi (termed “direct connections”) and, other than the cervical sacs, also have connections to the tertiary parabronchial network (termed “indirect”; see term history in Baumel et al., 1993; see also Lawson et al., 2021).

FIGURE 1.

Air sacs and bronchial tree morphology in Psittacus erithacus. 3D surface models of the pulmonary air sacs from μCT scans of P. erithacus (Aristotle). (a) Left lateral view of the postcranial axial skeleton and air sacs; (b) left craniolateral view of air sacs with the selection box demonstrating secondary bronchi connecting to cranial thoracic (V3), caudal thoracic (L3), and abdominal (Pb) air sacs; (c) left lateral view of semi‐transparent lung revealing bronchial tree (skeletal elements removed); (d) left lateral view of bronchial tree with select secondary bronchi labeled; (e) caudal view of bronchial tree with connection sites between bronchi and air sacs labeled; and (f) dorsal view of bronchial tree with ventrobronchi and first dorsobronchus labeled. Ab, abdominal air sac; C1, first cervical vertebra; Ca, caudal thoracic air sac; Cr, cranial thoracic air sac; D2, second dorsobronchus; D7, seventh dorsobronchus; H, humerus; IC+C, combined model of interclavicular air sac and cervical air sacs; L3, third laterobronchus; L3‐Ca, connection site between third laterobronchus and caudal thoracic air sac; Pb, primary bronchus; Pb‐Ab, connection site between primary bronchus and abdominal air sac; T1, first thoracic vertebra; T4, fourth thoracic (dorsal) vertebra; V1, first ventrobronchus; V2, second ventrobronchus; V3, third ventrobronchus; V3‐Cr, connection site between third ventrobronchus and cranial thoracic air sac; V4, fourth ventrobronchus.

The unpaired interclavicular sac and its diverticula (Figure 1) surround structures in the subpulmonary cavity, including the trachea, proximal primary bronchi, heart, and vessels. Axillary (Figure 2a) and suprahumeral diverticula (Figure 2c) are consistently present (Table 2). Subscapular diverticula (Figure 2d) are present in all but one of the eight specimens and, when present, vary in the extent to which they cover the deep surface of the scapulae. Air was interposed between the pectoralis and supracoracoideus muscles in two of the specimens with no clear connection to the other air sacs. In Zeno, the air is only interposed on the left side. As the presence of these airspaces could not be ruled out as artifacts of decay or inflation, they were not included in the final analysis. Medially, the interclavicular sac consistently includes a sternocardiac diverticulum (Figure 2d) on the midline interposed between the heart and sternum that extends caudoventrally and partially overlaps on each side with the ventral limits of the cranial thoracic sacs. Narrow bands of variably present soft tissue are visible in cross‐sectional views of these sacs that likely correspond to “tissue bridges” described by Duncker (1971, 1979) who further suggested that they penetrate the “impar median diverticulum” and attach to the sternum to support the oblique septum near the pericardium. These bridges produce a rougher, more corrugated appearance on the 3D surface models (Figures 3a–c and 4b,c) like those seen in Duncker's endocasts (Duncker, 1971). In Plato, the sternocardiac diverticulum extends caudoventrally past the caudal sternal margin and covers the anterior surfaces of the abdominal viscera alongside the abdominal air sacs (Figure 3d).

FIGURE 2.

Diverticula and pneumatization patterns in P. erithacus. (a–c) Left craniolateral views of pneumatization in the postcranial axial skeleton of Aristotle, while (d) is a left lateral view of Plato. (a) Postcranial axial skeleton with proximal appendicular components (i.e., coroacoids, scapulae, and left humerus). (b) Illustration of axial skeleton pneumatization and select appendicular skeletal elements. Grayscale thresholding was adaptively used to illustrate the extent of pneumatic invasion into the segmented axial skeleton (semi‐transparent). Coloration indicates the matching invading respiratory element. (c and d) Air sacs and select labeled air sac diverticula. Ax‐IC, axillary diverticulum of the interclavicular sac; Fm‐Ab, femoral diverticulum of the abdominal sac; H, humerus; It‐C, intertransversarial diverticula of cervical sacs; Pr‐Ab, perirenal diverticulum of the abdominal sac; Sh‐IC, suprahumeral diverticulum of the interclavicular sac; Sm‐C, supramedullary diverticulum of cervical sac; Ss‐IC, subscapular diverticulum of interclavicular sac; St‐IC, sternocardiac diverticulum of interclavicular sac.

TABLE 2.

Diverticula in the air sacs of deceased (n = 7) and live (n = 2) specimens P. erithacus.

	Ax‐IC	Fm‐Ab	Notes	Pr‐Ab	Size	Sh‐IC	Sm‐C	Ss‐IC	Notes	St‐IC	Notes
Aristophanes	+	+		+	Small	+	+	+		+	Fibrous
Aristotle	+	+		+	Small	+	+	−		+	Fibrous
Homer	+	−		+	Small	+	+	+	Double	+	Fibrous
Plato	+	+	Large	+	Large	+	+	+		+	Fibrous
Pythagoras	+	+	Fibrous	+	Large	+	+	+		+	Fibrous
Socrates	+	+		+	Small	+	+	+	Small	+	Fibrous
Zeno	+	+		+	Small	+	+	+		+	Fibrous
Diogenes*	+	+		+		+	+	+		+	Fibrous
Hesiod*	+	+		+		+	+	+		+	Fibrous

Note: Live birds are indicated with asterisks. Presence (+) or absence (−) indicate that diverticula are found or not found bilaterally. Select diverticula include a column to the right with additional descriptors of relative size of the diverticula or the presence of intervening “fibrous bridges” of soft tissue.

Abbreviations: Ax‐IC, axillary; Fm‐Ab, femoral; Pr‐Ab, perirenal; Sh‐IC, suprahumeral; Sm‐C, supramedullary; Ss‐IC, subscapular; St‐IC, sternocardiac.

FIGURE 3.

3D surface models of the air sacs in P. erithacus (n = 4/7). Left lateral (left column) and ventral (right column) views of (a) Aristophanes; (b) Aristotle; (c) Homer; and (d) Plato.

FIGURE 4.

3D surface models of the air sacs in P. erithacus (n = 3/7). Left lateral (left column) and ventral (right column) views of (a) Pythagoras; (b) Socrates; and (c) Zeno.

The cervical sacs emerge from a branch of the first ventrobronchus and extend cranially and caudally, bearing interconnected diverticula that course through the vertebral canal, transverse foramina (Figure 2c), and along the external surfaces of the vertebrae (“supramedullary,” “intertransversarial,” and “supravertebral diverticula,” respectively in Müller, 1908; for a recent review of these diverticula, see Atterholt & Wedel, 2023). The supramedullary diverticulum (Figure 2c) also accompanies all the thoracic vertebrae and cranial portions of the synsacrum, occasionally apposing the tissue surfaces of the bronchi coursing along the outer margins of the gas‐exchanging lung. Fine tissue planes are variably observable between the supramedullary diverticula and the bronchial network of the lung, and whether there is communication between these extensions of the cervical sac and the conduits of the gas‐exchanging lung is unclear. Intertransversarial diverticula were partially inflated but visible in transverse foramina throughout the cervical region in all scans. Supravertebral diverticula are uninflated or mostly absent at the thoracic level but may appear more caudally after the supramedullary diverticulum has terminated within the synsacral vertebral canal. The origins of some observed supravertebral diverticula are indistinct due to the presence of thin tissue bridges, and they may variably represent extensions of cervical air sacs, abdominal air sacs, or discrete saccular extensions of nearby bronchial conduits. In vivo scans (Figure 5) exhibit similar features and diverticula for both the interclavicular and cervical sacs, albeit less inflated. Hesiod exhibits an asymmetric sternocardiac diverticulum on the left ventral surface (Figure 5b) but it is unclear if this is a consequence of partial inflation or an anatomical variation.

FIGURE 5.

Air sacs in vivo in P. erithacus. 3D surface models of the air sacs in vivo from P. erithacus (n = 2). Left lateral (left side) and ventral (right side) views of (a) Diogenes and (b) Hesiod. C‐Cr, costal impressions visible on left cranial thoracic air sac; Pr‐Ab, partially deflated perirenal diverticulum of the left abdominal air sac; St‐IC, sternocardiac diverticulum of the interclavicular sac.

Compared with the other air sacs observed in P. erithacus, the cranial thoracic sacs are bilaterally symmetric and intraspecifically consistent. Sternal ribs are intimately apposed to these sacs and distinct impressions of the ribs are visible as furrows on the ventrolateral surfaces of the models (Figures 3 and 4). They arise from a branch of the third ventrobronchus and form indirect connections with the parabronchial network that have, in some accounts, been described as “recurrent bronchi” (Locy & Larsell, 1916b).

The caudal thoracic sacs arise from large laterobronchi and occupy a position caudolateral to the gas‐exchanging lung. They are variably sized in the specimens relative to the other air sacs (Figures 3 and 4) and, consequently, have variable relationships to the cranial thoracic and abdominal air sacs along their margins. In some specimens, for example, some caudal thoracic air sacs are large enough to have a free ventral margin (Figures 3a and 4b) that in others are apposed to the cranial thoracic sac (Figures 3b,c, 4a,c and 5). They are often asymmetric (Figure 6b). In Aristophanes (Figure 3a), the left sac is much smaller, and an asymmetric absence of the radiopaque soft tissue boundaries along the medial edge indicates that it has fused with the immediately adjacent abdominal air sac (Figure 7). The boundary, which is clearly visible between the right sacs (Figure 7d), would typically include the oblique septum that sequesters the abdominal sac to the intestinal part of the coelom. In vivo scans exhibit partially inflated cranial and caudal thoracic sacs but still exhibit clear rib impressions on their lateral surfaces (Figure 5).

FIGURE 6.

Asymmetric air sac examples. Ventral views of the primary bronchi connected air sacs in Zeno (a, b) and Socrates (c). Secondary bronchi have been separated and connected by a dotted line to reveal the entire ventral surfaces of abdominal air sacs. L‐Ab, Left abdominal air sac; L‐Ca, Left caudal thoracic air sac; R‐Ab, Right abdominal air sac.

FIGURE 7.

3D surface model of the fused air sacs in P. erithacus (Aristophanes) with select bronchi and semi‐transparent lung surface. (a) Left lateral view of fused sacs and their direct connections to the bronchial tree and (b) left craniodorsal view of the fused left air sac and corresponding unfused right air sacs with the semi‐transparent lung removed. Vertical lines indicate axial sections in the panels below. (c & d) Sites of fusion between left abdominal and caudal thoracic air sacs. Arrowheads are in the left and right abdominal air sacs. Closed pink arrowheads point at the intact soft tissue plane separating the right abdominal and caudal thoracic air sacs. Open pink arrowheads point at soft tissue planes surrounding the fusion site between the left abdominal and caudal thoracic air sac. (e) View of intact tissue planes caudal to the left fusion site. Asterisks in panels (c–e) indicate the region corresponding to a fused rudimentary left caudal thoracic air sac. Ab, right abdominal air sac; Ca, right caudal thoracic air sac; Cr, right cranial thoracic air sac; Fu, fused left caudal thoracic and abdominal air sacs; Pb, right primary bronchus.

The abdominal sacs extend through the horizontal septum into the intestinal subdivision of the coelom (“intestinal peritoneal cavity,” Taylor, 2016) to variably surround the intestines, kidneys, and reproductive organs (Figure 6a,c). In all examined birds, the abdominal sacs include the perirenal diverticula (Figure 2c) extending craniodorsally to occupy the renal fossa. All birds except Homer have femoral diverticula (Table 2). The abdominal air sacs are usually asymmetric, surrounding the digestive organs differently in each bird. In Socrates, the right abdominal sac is much larger than the left, occupying a substantial caudal portion of the coelom by itself when fully inflated (Figure 6c). In contrast to the artificially inflated air sacs of the deceased specimens, in vivo scans exhibit partially deflated portions of the abdominal air sacs, although the perirenal diverticula and the proximal portion of the sacs that are in direct connection with the primary bronchus are still visible (Figure 5).

3.2. Variation in skeletal pneumatization patterns

All five specimens examined exhibit relatively consistent patterns in terms of which skeletal elements are pneumatized. The locations of the pneumatic foramina vary regionally and intraspecifically (Figure 8). Based on the criterion of a pneumatic foramen being a break in cortical bone connecting an air‐filled respiratory element with an air‐filled skeletal region, the analysis identified hundreds of unambiguous foramina on each skeleton, ranging from 168 in Aristotle to 243 in Aristophanes (Table S1). Checks for inter‐observer reliability in Aristotle and Homer indicated occasional variation in the authors ABL and AM locating, identifying, and counting smaller (~100–200 μm) foramina, so analysis is limited to broad descriptive comparisons of foramen locations on skeletal surfaces rather than quantitative comparisons of foramen totals. The scan of Plato was selected for quality (see Methods) but the image stack was truncated at its cranial extent (see Table 1), so the comparison of all the cervical vertebrae was limited to the remaining four birds.

FIGURE 8.

Pneumatization location and symmetry patterns. (a–f) Illustrate how often a specific pneumatization pattern is seen in the μCT scans of P. erithacus (n = 5). Purple indicates bones that were apneumatic. Gray indicates bones for which a particular comparison was not applicable. The strength of coloration in each panel indicates how many of the specimens exhibited a particular feature for a particular midline or paired pneumatized skeletal element (e.g., the lightest green indicates that one of the examined birds exhibited a specific pneumatic feature for a single vertebra or for a pair of ribs). Pneumatic features assessed for each element include (a) which elements were pneumatized and (b) which single or paired bones exhibited asymmetric locations for pneumatic foramina. Panels (c–e) indicate the number of birds that had pneumatic foramina invaded from (c) the vertebral canal; (d) the lateral surface of the centrum; or (e) from the transverse foramina. Panel (f) indicates the presence of pneumatic foramina in the transverse processes. C3, third cervical vertebra; C11, eleventh caudal vertebra; Ca1, first caudal vertebtra; Ca5, fifth caudal vertebra; Co, coracoids; Fr, furcula; Pv, pelvis; Sc, scapulae; Sy, synsacrum; T1, first thoracic vertebra; T4, fourth thoracic vertebra; T8, eighth thoracic vertebra.

3.2.1. Postcranial axial skeleton

The axial skeleton of P. erithacus includes a total of 11 cervical vertebrae, eight thoracic vertebrae, the synsacrum, the caudal vertebrae, the pygostyle, the sternum, and costal elements that either fuse or articulate with the vertebrae depending on the region (Figure 8). Cervical vertebrae have bilateral costal elements that fuse with the vertebral body and thereby contribute to the ring of bone (“ansae costotransversarii,” Baumel et al., 1983) surrounding each transverse foramen. In the thoracic region, costal elements are instead represented by forked vertebral ribs that articulate with an adjacent thoracic (or “dorsal”) vertebra, and each side of these vertebrae includes a bony diapophysis (i.e., transverse process) and parapophysis (i.e., costolateral eminence) that articulate with the respective tubercle and capitulum of an associated vertebral rib. Each rib and its corresponding vertebra together consequently enclose a space that is morphologically equivalent to the transverse foramen of the cervical vertebrae (Baumel et al., 1993). Pneumatic foramina are most typically located near these equivalent spaces where vertebral and costal skeletal elements fuse or articulate, although foramina are also occasionally found on the bony surfaces lining the vertebral canal (Figure 9). Pneumatic foramina on the dorsal surface of the vertebral arches or the ventral portions of the centra are relatively rare. The atlas (C1), caudal vertebrae, and pygostyle are apneumatic in all birds examined (Figure 8).

FIGURE 9.

Examples of variable pneumatic foramen locations. Caudoventral view of surface models (a, b) and schematics (c, d) for the C8 vertebra in Aristotle (a, c) and Plato (b, d). Pink regions indicate sites of pneumatic foramina variably located on the centrum, ansa costotransversaria, and vertebral arch. An, right ansa costotransversaria; V, vertebral arch.

Cervical vertebrae (C1–C11)

Except for the consistently apneumatic atlas (C1), cervical vertebrae are typically pneumatized on the inner bony surfaces surrounding the transverse foramina (Figures 8e and 9). As previously mentioned, C1 is consistently apneumatic in all specimens. The axis (C2) is only apneumatic in Aristophanes and Homer. Pneumatic foramina of these skeletal elements are connected to diverticula of the cervical sacs including (1) bilateral canalicular diverticula coursing through the transverse foramina with accompanying neurovasculature (called “intertransversarial canals,” Müller, 1908) and (2) a midline diverticulum accompanying the spinal cord in the vertebral canal (“supramedullary diverticulum,” Müller, 1908). In Aristotle, the diameter of C2–C5 pneumatic foramina are often at the limits of the scan resolution (~100 μm) and only traceable as 2–3 pixels of radiolucent air crossing a break in the cortical surface on sequential reconstructed planar slices. The second cervical vertebra in Aristotle was completely pneumatized, but a pneumatic foramen could not be unambiguously located on its cortical surface (Table S1). C3–C11 are primarily pneumatized from the intertransversarial canals by pneumatic foramina that are variably positioned on the inner surfaces of the ansae costotransversarii (Figure 8e). Pneumatic foramina on the lateral surface of the centra and pedicles are more common near the cervicodorsal transition from cervical to thoracic vertebrae (Figure 8d).

Thoracic vertebrae (T1–T8)

The thoracic series is associated with mobile costal elements, and in P. erithacus includes the transitionary T1–T3 “cervicodorsal” vertebrae (Baumel et al., 1993) at the boundary between the cervical and thoracic series. T1 articulates with rudimentary vertebral ribs resembling the fused ansae costotransversarii of the cervical series. T2 and T3 articulate with forked vertebral ribs that lack a sternal counterpart. The remaining T4–T8 vertebrae have been described as “complete” or “true” ribs because they form a connection with the sternum via both a vertebral and sternal rib element (Baumel et al., 1993).

Thoracic vertebrae are almost always pneumatized directly by the gas‐exchanging lung via relatively small pneumatic foramina on the ventromedial, craniomedial, and lateral tips of the transverse processes (Figure 8f) or on the lateral centrum (Figure 8d). T1 and T2 lack pneumatic foramina invading from diverticula in the vertebral canal (Figure 8c). T2 is variably pneumatized by both the cervical air sacs and pneumatizing diverticula emerging directly from the gas‐exchanging lung in Aristotle but is exclusively pneumatized by the gas‐exchanging lung in Plato. Pneumatic foramina are on the bony surfaces forming the inner lining of the vertebral arches via the supramedullary diverticula (Figure 2c). While most of these foramina extend from the diverticulum dorsally or laterally into the vertebral arch, they also occasionally extend ventrally into the thoracic vertebral centra. A unique, asymmetric, and unambiguous foramen was seen in Plato connecting a ventrobronchus of the right lung to the pneumatized space of the relatively long and distinct ventral spine (hypapophysis) of T4. Foramina from the supramedullary diverticulum are more typically located in the caudalmost thoracic vertebrae and synsacrum (Figure 8c).

Synsacrum

In all five birds, the synsacrum exhibits pneumatic foramina on the cranialmost fused synsacral vertebrae (S1 and S2). Notably, these vertebrae are also variably associated with free vertebral rib elements (Figure 8). Bilateral renal fossae that form part of the ventral synsacral surface represent the fusion of the synsacral transverse processes with the ilium and, in the birds examined, included pneumatic foramina from the perirenal diverticula (Figure 2c) of the abdominal air sacs. Diverticula near the caudal boundary of the gas‐exchanging lungs are variably present and have variable pneumatization patterns. Notably, these gas‐exchanging lung‐derived diverticula are entirely absent in Aristotle while, in Plato, they extend far enough to pneumatize the synsacrum. In the remaining birds, these extensions exhibit relatively ambiguous boundaries in the μCT images that obscure whether they are diverticula derived from the lung or from the air sacs.

Sternum

The interclavicular sac pneumatizes the sternum on its internal bony surface via large bilateral pneumatic foramina near the coracoid articulations. Additional pneumatic foramina of the sternum can be identified on its dorsal (i.e., deep) surface as invading from the cranial thoracic sacs. Although often ambiguous due to the presence of fluid, pneumatic foramina in this region are typically located near the sternal articulations with sternal ribs, and invasion at these sites pneumatize the bilateral columns of sternal bone (“pila costalis,” Baumel et al., 1993) to form a contiguous internal column of air. The remaining dorsal pneumatic foramina are variably present and invade the sternum asymmetrically (Figure 8b).

Ribs (T1–S2)

The following describes pneumatization of rib elements associated with the cervicodorsal vertebrae (T1–T3), “true” thoracic vertebrae (T4–T8), and synsacral vertebrae (S1 and S2). Almost all of these rib elements exhibit pneumatic foramina near their articulations with either the vertebral column or the sternum, although there are a few exceptions (Figure 8a). For example, the right T6 vertebral rib is apneumatic in Aristotle. All birds except Zeno have a pair of pneumatized vertebral ribs that articulate with the S1 vertebra of the synsacrum. Additional rib elements associated with the synsacrum are variably present (e.g., S1 sternal ribs) and sometimes pneumatized (Table S1).

Vertebral ribs are typically pneumatized medially where they articulate with a corresponding vertebra via forked processes. Motion artifacts and fluid accumulation were common in the scans of deceased birds near these vertebral rib articulations and may in part be explained by (1) the dorsal recumbent position of all birds but Plato, (2) the beginning of slight decay in the frozen specimens, and/or (3) the accumulation of condensation from thawing. Regardless, pneumatic foramina of the vertebral ribs were observed to be consistently located on the surfaces of the ventrally positioned capitulum and dorsally positioned tubercle as the ribs articulate at their medial extent with the parapophysis and diapophysis of a vertebra, respectively. Pneumatic foramina are also seen where the processes join at the neck of the rib (“capitulotubercular notch,” Baumel et al., 1983). While cervicodorsal vertebral ribs (T1–T3) are typically pneumatized by the cervical sacs, T4–T8 and S1 vertebral ribs are almost entirely pneumatized directly by diverticula from the gas‐exchanging lung. In Aristophanes and Zeno, T3 vertebral ribs are principally pneumatized by the cervical sacs. In Aristotle, the right T6 vertebral rib is apneumatic, and the left T4 vertebral rib includes a pneumatic foramen connecting with the surface of the cranial thoracic sac near the articulation with the T4 sternal rib. The left S1 vertebral rib in Homer is apneumatic, while the right is pneumatized by the abdominal sac. S1 vertebral ribs are bilaterally pneumatized in Plato and Zeno, however, pneumatic foramina could not be identified on the left side in either specimen. The bronchial airways pneumatize the right S1 vertebral rib in Plato, but the same rib in Zeno is pneumatized near its articulation with the S1 sternal rib by the caudal thoracic sac.

Rudimentary ribs are also variably found associated with T1 and, in Aristotle, with the S2 segment of the synsacrum. The rudimentary ribs articulating with the T1 vertebra are pneumatized by the cervical air sac bilaterally in Plato, unilaterally on the left side in Homer (Table S1) and are apneumatic in the three remaining birds. There is also a pair of rudimentary S2 ribs in Aristotle, the right of which only articulates at its tubercle and is pneumatized by the abdominal air sac. The corresponding left rudimentary S2 rib is non‐articular, embedded entirely in the soft tissues of the body wall, and is apneumatic.

The sternal ribs articulate with the distal ends of the T4–S1 vertebral ribs dorsally and, ventrally, with costal processes on the lateral margin of the sternum. The corresponding pneumatic foramina of these ribs can be found at their ventral extent on their cranial or caudal surface as they articulate with the sternum, although a few of the ribs are small enough that no obvious foramina are visible. The S1 sternal rib is present in all birds but is not typically pneumatized (Figure 8a). Homer, Plato, and Zeno have additional, apneumatic sternal ribs that variably articulate with the T2 vertebral ribs, but not the sternum. Sternal ribs are invaded by cranial thoracic sacs, although these findings are confounded by variably present tissue bridges associated with the adjacent sternocardiac diverticulum of the interclavicular sac.

3.2.2. Shoulder and hip bones

Proximal skeletal elements of the pectoral girdle and shoulder are similarly pneumatized in all birds (Figure 8a) by extensions of the interclavicular sac but occasionally exhibit asymmetric foramina locations (Figure 8b). Humeri are pneumatized at the pneumotricipital fossa in all birds. Coracoids are pneumatized at the cranial edge of the supracoracoid sulcus near the articulation with the scapulae, but Aristotle exhibits an additional pneumatic foramen that is asymmetrically present near the right humeral facet. These same interclavicular sac extensions additionally pneumatize scapulae in the other birds via foramina on the acromion, although the scapulae in Aristotle are not pneumatized. The furcula and remaining forelimb bones are apneumatic in all specimens examined.

The partially fused “Os coxae” (ilia, ischia, pubes) form the bony pelvis with the synsacrum. Large pneumatic foramina are invaded by the perirenal diverticula from the abdominal air sacs at or near the cranial portion of the renal fossa. There are variably present and often asymmetric pneumatic foramina (Figure 8b) found more cranially along the ventral surface of the preacetabular alae or from femoral diverticula of the abdominal air sac that pneumatize the dorsal ilium near the preacetabular tubercles. The ischium and pubis lack these foramina, and so are presumedly pneumatized via expansion across the synostosis with the ilium. All hindlimb bones distal to and including the femur are apneumatic.

3.2.3. Symmetry/asymmetry

Although most of the pneumatized bones exhibit symmetrical placements of pneumatic foramina, exceptions are identifiable throughout, and these asymmetric sites are more often located on the thoracic vertebrae, sterna, and pelvic bones (Figure 8b). Thoracic vertebrae exhibit asymmetric foramina located on the pedicles of the vertebral arches, pneumatizing either from the lateral surfaces of the arch (Figure 8d) or from the vertebral foramen (Figure 8c). Pneumatic foramina on the pelvis are typically observed along the renal fossa or pre‐acetabular ala of the ilium but include many exceptions in the examined birds, including (1) an asymmetric foramen located on the right ilioischiadic suture in Zeno; (2) an asymmetric foramen on the pre‐acetabular ala in Aristophanes, Aristotle, and Homer; and (3) an asymmetric foramen on the renal fossa in Homer and Plato. In all birds, sterna are consistently invaded on their internal surfaces via large, cranially positioned foramina but may also be variably pneumatized near the coracoid pillars, costal pillars, costal incisures, or along other regions of the internal surface (e.g., pars cardiaca, pars hepatica, or the craniolateral processes).

The cervical vertebrae, ribs, and appendicular bones exhibit relatively fewer asymmetric sites of pneumatization (Figure 8b). Cervical vertebrae typically exhibit bilateral pneumatic foramina on the inner surfaces of the ansae costotransversarii (Figure 8e) but are often accompanied by unpaired pneumatic foramina on the inner surfaces of the transverse foramina, within the vertebral canal, or on the lateral surface of the centrum (Figure 9). Ribs are typically invaded on both sides near their respective vertebral and sternal articulations. Appendicular elements have large, consistent pneumatic foramina associated with specific bony landmarks.

4. DISCUSSION

Analysis of the air sacs in deceased and live scans of P. erithacus reveals (1) low intraspecific variation of air sac positioning; (2) consistent connections to the bronchial tree; (3) variable origin and extent of their many diverticula; and (4) individual specimens with asymmetric caudal thoracic and abdominal air sacs. Assessment of the pneumatization patterns in the deceased birds reveals (5) regional variation in the location of pneumatic foramina; (6) asymmetric pneumatic foramen locations throughout the postcranial skeleton that are often found on the thoracic vertebrae, sternum, or pelvis; and (7) apneumatic caudal vertebrae, pygostyle, and bones that make up the remaining appendicular skeleton.

4.1. Air sac variation and asymmetry

The relative consistency of air sac positioning in P. erithacus and variability of the origin and extent of the femoral and perirenal diverticula suggests that the expansion of air sacs to surround adjacent tissues is partially opportunistic and that this aspect of the adult respiratory morphology exhibits intraspecific anatomical variation similar to other distributive organ systems (e.g., nerves and blood vessels). A speculative explanation for the diminutive caudal thoracic air sacs in Aristophanes and Socrates, for example, is that cranial thoracic and abdominal air sacs grew slightly faster post‐hatching, then overtook and limited continued expansion of the caudal thoracic sacs. Similar variations in timing could explain asymmetries observed in the abdominal sacs of Socrates, the caudal thoracic sacs of Aristophanes, and the extensive variation in the length and width of the sternocardiac diverticulum in all specimens (Figures 3 and 4). Additionally, this may have resulted from pathological disturbances in the blood supply during development, or from prominent scoliosis present in Socrates (diagnosed by author MSE). Further experimental testing would be needed to confirm when and how these variations emerge.

The fusion of the left caudal thoracic and abdominal air sacs observed in scans of Aristophanes indicate (1) that intraspecific variations may include air sac fusion and (2) that the lungs of P. erithacus can functionally accommodate this asymmetry. This fusion pattern is also mentioned, though without indication of asymmetry, as occurring in the mute swan (Cygnus olor) by Duncker (1971). While additional air sac fusions have been described across taxa during development for the unpaired interclavicular sac (Locy & Larsell, 1916b), between interclavicular and cranial thoracic sacs in songbirds (Duncker, 1971) and in a variety of sacs for specific taxa (King, 1966), the fusion observed here and previously by Duncker suggests asymmetric absence or rupture of the oblique septum that typically sequesters the abdominal sacs in the intestinal part of the coelom. This unusual finding suggests that loss of separation between the caudal thoracic and abdominal air sacs by the oblique septa may be compatible with functional respiration.

The overall similarity of air sac arrangements in P. erithacus that can also accommodate variations (e.g., relative size variations or fusions) suggests that individual air sacs may not be functionally constrained to occupy specific positions or relationships to one another. While the previously mentioned findings of fusions in other taxa do not necessarily indicate broader cranial‐caudal modularity for avian air sacs, the apparent tendency of caudally positioned air sacs (e.g., caudal thoracic and abdominal sacs) to exhibit the most obvious gross variations in P. erithacus indicates a need for further study. For example, quantitative comparisons of dynamic respiratory air sac volumes (e.g., tidal, reserve, and vital capacities) in this and other taxa may reveal interspecific differences that underpin functional subdivisions of the air‐sac system more generally.

4.2. Pneumatization patterns

Pneumatic foramen locations observed in P. erithacus reveal extensive and regionally variable pneumatization of axial skeletal elements, although these foramina were located most consistently at the boundary between midline (e.g., vertebrae and sterna) and costal (e.g., ribs) bony elements (Figure 8e,f). Foramina located on the bony surfaces of the vertebral canal (Figure 8c) and lateral centra (Figure 8d) are more common in the cranialmost thoracic vertebrae. Preferential localization of foramina on bony lines of fusion (e.g., vertebral/costal and pelvic suture lines) transitionary regions (e.g., cervicodorsal vertebrae), and vascular pathways (e.g., transverse foramina) suggest that pneumatic foramina may be typically formed at the margins of developmental territories and, like other vertebrate body systems (e.g., circulatory, peripheral nervous, and integumentary), may be prone to an anatomical variation in size, location, and number. If future research finds correlations between the vascular, skeletal, and respiratory anatomy in pre‐ and post‐hatch birds, it would likely have less impact on functional hypotheses related to overall changes in body size (e.g., weight distribution and buoyancy in Smith, 2012) but may broadly explain regional pneumatization patterns observed in taxa (e.g., as in the centrum‐first/arch‐first pattern observed in non‐avian theropods by Benson et al., 2012) as indicative of vascular—rather than respiratory—homology.

Overall, analysis of the lower respiratory system in whole bird specimens of P. erithacus matches with what has been described as the most common arrangement for air sacs (Duncker, 1971; King, 1966; McLelland, 1989) and pneumatization extent (Baer, 1896 via King, 1966). The observed intraspecific variation suggests some flexibility in their arrangement for this taxon. Postcranial skeletal elements are variably pneumatized by adjacent air sacs, pulmonary diverticula, and/or directly by the airways of the lung. These findings contradict widely cited publications attributing all pneumatization exclusively to the air sacs (Bremer, 1940b; Hogg, 1984b; King, 1966; King & Kelly, 1956) and align with more recent studies of pneumatization (Atterholt & Wedel, 2023; O'Connor, 2006; Schachner et al., 2021). Pneumatic foramen locations for pneumatized portions of the appendicular skeleton are consistent and symmetric relative to the axial skeleton, particularly when compared to the vertebral column, sternum, and pelvis.

Pattern variations for these vertebral elements suggest that the arrangement of air sacs and the location of pneumatic foramina for these bones reflect morphologies analogous to other systems that distribute resources (e.g., blood and innervation) from a centralized set of organs. Air sac arrangement, diverticulum extent, and pneumatic foramen location—rather than indicating any functional role they play in gas‐exchange or ventilation—may represent the variable alignment of blood supply and lung tissues that provided the most expedient means for distributing air and potential air‐filled spaces to bones, tissues, or partitions of the avian coelom. The more consistent arrangement of air sacs in the thorax and associated ribcage pneumatization may indicate either (1) a functional constraint imposed by the respiratory system to couple thoracic air sac volume changes with rib movement or (2) less variability in the circumstantial developmental factors leading to pneumatization. Vertebral ribs, for example, may be pneumatized in a more consistent pattern than the adjacent thoracic vertebrae due to the reduced likelihood of pneumatization involving the vascular networks of the vertebral canal. Analysis of additional taxa would be needed to confirm this hypothesis.

The observed intraspecific variation in air sac and pneumatization patterns for P. erithacus indicate a need to establish methodological baselines for identifying variation in postcranial pulmonary structures, determining which bones are pneumatized, and quantifying the contributions of intraspecific anatomical variation to postcranial pulmonary variation more generally. O'Connor developed strategies to meet these needs by comparing simple ratios of pneumatized skeletal elements to the entire skeleton (2004) and by developing precise terminology for defining and distinguishing pneumatic foramina (2006). Using a mix of ex vivo bone specimens and endocasts, these criteria were then employed to identify variation within anseriform clades and other taxa (2009) to describe “common” and “extended” pneumatization patterns dependent on size, locomotion, and foraging specializations (Gutherz & O'Connor, 2021). Foramina with a diameter less than 500 microns, however, were excluded due to the impracticality of minute identifications on dry bone, and the methods relied additionally on other features (e.g., number, size, and appearance) to distinguish between pneumatic and exclusively vascular foramina (O'Connor, 2006). In line with Wedel's suggestion (via Apostolaki et al., 2015), the pneumatic foramina identified here in P. erithacus using μCT data indicate that pneumatic foramina may regularly be much smaller than 500 microns and may be indistinguishable ex vivo from other neurovascular passages. Recent studies comparing medullary bone deposition in several galloanseriform and neoavian taxa (Canoville et al., 2019) include findings that incidentally conflict with O'Connor's work on pneumatization patterns, suggesting a need for studies that can refine the existing criterion for evaluating pneumatization and the subsequent hypotheses they inform.

4.3. Impacts of imaging protocols and specimen condition

The combined findings from living and deceased whole bird scan data reveal variability in the arrangement of air sacs that may in part reflect differences in how the individual birds were scanned and their unique body conditions. Marked differences observed in the volume of abdominal air sacs between deceased and live specimens, therefore, may be explained by (1) shallow breathing under general anesthetic, (2) compression of the air sacs by adjacent abdominal viscera (Malka et al., 2009; Nevitt et al., 2014), and/or (3) prior or current underlying disease that may have contributed to anatomic alterations. Additionally, all the deceased specimens were artificially inflated to their maximum inspiratory capacity which facilitates visualizing the pulmonary structures in deceased taxa but likely overestimates the typical sizes of the sacs in eupnea.

While the imaging data and analysis presented here reveal new levels of detail for avian anatomy, μCT introduces new artifacts and challenges for interpretation. In addition to scanning and positioning artifacts, μCT relies on algorithms that potentially add their own subtle artifacts to data and, regardless of quality, cannot be used in many cases to distinguish between soft tissues with similar radiodensity (e.g., blood, fluid, or muscle). As such, ambiguous foramina (that were excluded from the analysis here) may resolve in other data types (e.g., endocasts, plain radiographs, or dissections) to be vasculature channels or fluid‐filled respiratory spaces. As such, confirmation of these findings should employ variations to these imaging approaches that can be corroborated with these and more traditional methods.

While records for each specimen indicated that no respiratory disease was present, the image data exhibit indications of acquired health issues typical of captive psittacines. Notably, Socrates had visible signs of scoliosis affecting the thoracic region that may have contributed to the unique air sac asymmetries and the air sac fusion observed in this specimen. All skeletons of the live and deceased birds additionally showed signs of osteopenia. The cause of death was not described for the deceased specimens and additional disorders may have been present. While the variable conditions of specimens prevented any meaningful quantitative analysis of bone density, trabecular density, or air sac volumes, observed pneumatic foramen locations and overall air sac morphologies were assumed, as the literature indicates (Bremer, 1940b; Hogg, 1984a), to have been driven by conditions preceding their observed approximate 2‐month growth period post‐hatching.

5. CONCLUSIONS

The lower respiratory system of P. erithacus can accommodate substantial variation and asymmetry in the arrangement of the air sacs, and midline skeletal elements exhibit greater variability in pneumatic foramen location relative to the pneumatized costal and appendicular elements. Notable variations in air sacs include (1) variably sized sternocardiac diverticula of the interclavicular sac along the internal surface of the sternum; (2) asymmetrically‐sized abdominal air sacs that filled a substantial portion of the entire caudal coelom; and (3) a diminutive left caudal air sac in Aristophanes that appears fused to the abdominal sac. In vivo scans additionally reveal that the abdominal sac is capable of nearly complete deflation during sedated respiration. Together, these findings support the hypotheses that air sac positioning and pneumatic foramen location in the adult are reflective of opportunistic interactions between soft tissue, blood supply, and bone during their development.

AUTHOR CONTRIBUTIONS

Adam B. Lawson: concept/design; data curation; analysis; interpretation; methodology; project administration; software; validation; visualization; manuscript drafting; critical revision of manuscript. Aracely Martinez: data curation; validation; critical revision of the manuscript. Brandon P. Hedrick: concept/design; analysis; methodology; supervision; critical revision of manuscript. M. Scott Echols: concept/design, data curation; funding acquisition; methodology; resources; critical revision of the manuscript. Emma R. Schachner: concept/design; data curation; analysis; funding acquisition; methodology; project administration; resources; supervision; validation, visualization; critical revision of the manuscript.

FUNDING INFORMATION

Association of Avian Veterinarians (AAV Research Grant) to ERS and MSE.

CONFLICT OF INTEREST STATEMENT

The authors have no conflict of interest to declare.

Supporting information

Table S1.

Lawson, A.B. , Martinez, A. , Hedrick, B.P. , Echols, M.S. & Schachner, E.R. (2025) Variation in air sac morphology and postcranial skeletal pneumatization patterns in the African grey parrot. Journal of Anatomy, 246, 1–19. Available from: 10.1111/joa.14146

DATA AVAILABILITY STATEMENT

The manuscript includes this statement and direct link to the data:All imaging data are available via MorphoSource in DICOM or TIFF format: https://www.morphosource.org/projects/000554139.