Avian palaeoneurology: Reflections on the eve of its 200th anniversary

Abstract

In birds, the brain (especially the telencephalon) is remarkably developed, both in relative volume and complexity. Unlike in most early‐branching sauropsids, the adults of birds and other archosaurs have a well‐ossified neurocranium. In contrast to the situation in most of their reptilian relatives but similar to what can be seen in mammals, the brains of birds fit closely to the endocranial cavity so that their major external features are reflected in the endocasts. This makes birds a highly suitable group for palaeoneurological investigations. The first observation about the brain in a long‐extinct bird was made in the first quarter of the 19th century. However, it was not until the 2000s and the application of modern imaging technologies that avian palaeoneurology really took off. Understanding how the mode of life is reflected in the external morphology of the brains of birds is but one of several future directions in which avian palaeoneurological research may extend. Although the number of fossil specimens suitable for palaeoneurological explorations is considerably smaller in birds than in mammals and will very likely remain so, the coming years will certainly witness a momentous strengthening of this rapidly growing field of research at the overlap between ornithology, palaeontology, evolutionary biology and neurosciences.

In anticipation of the 200th anniversary of the first published mention of an endocast of the brain cavity in a fossil bird, the authors review the history and approaches of avian palaeoneurological investigations and offer some perspectives on where this rapidly growing field at the overlap between ornithology, palaeontology, evolutionary biology and neurosciences should be going.

1. HISTORICAL BACKGROUND

Avian palaeoneurology has been a relatively neglected field of research (for earlier reviews that either centred on or included avian palaeoneurology see Walsh and Milner, 2011a; Walsh and Knoll;2011; Walsh and Knoll, 2018). This is no doubt due to the fact that the skulls of birds are particularly fragile and, therefore, not commonly found in fossil sites (Gardner et al., 2016). Even when the skull is preserved, it may be crushed, limiting considerably its potential for palaeoneurological investigations. Thus, the vast majority of Early Cretaceous birds that could theoretically be studied from the viewpoint of palaeoneurology, such as those of the famous Jehol biota, have flattened skeletons (see Wang et al., 2016). Yet, despite these difficult circumstances, palaeoneurological investigations on birds have a long history (Figure 1).

Figure 1.

Timeline of milestones in avian palaeoneurology. Only the works that are deemed influential in some way in this field are shown, irrespective of whether they were actually about the study of endocasts of fossil birds or not. The time frame covered begins with Cuvier (1804) and ends with Beyrand et al. (2019)

2. PIONEERING DAYS

The first mention of an avian fossil brain is to be found in a nota bene in the second edition of Cuvier’s (1822: 328) Recherches sur les Ossemens fossiles, a seminal work in vertebrate palaeontology. The discreet text reads: ‘Au moment où l’on achève l’impression de cette feuille, je reçois encore de Montmartre, un Ornitholithe, où la tête, le cou, l’aile, le croupion, la cuisse, et ce qui est plus extraordinaire, la trachée‐artère, sont en place et bien conservés; on y distingue jusqu’aux osselets qui renforcent la sclérotique, et jusqu’à l’empreinte du cerveau’ [At the moment when the printing of this publication is being achieved, I receive again from Montmartre, an Ornitholithe, in which the head, neck, wing, rump, thigh, and, what is more extraordinary, the windpipe, are in place and well preserved; even the ossicles that reinforce the sclera and the impression of the brain can be distinguished]. The specimen, which indeed shows a natural endocast, would never be described by Cuvier.

Cuvier was certainly not surprised by the fact that the brain morphology could be observed in this particular fossil. In fact, he had already come across several natural endocranial casts during the course of his studies of the extinct mammals from the Eocene of Montmartre (Cuvier, 1822: 37, 38, 44), a commune subsequently annexed to Paris, France. He first described and figured one of these endocasts at the very beginning of the 19th century (Cuvier, 1804: 300‐301, pl. VII fig. 3), thereby building the foundation of palaeoneurology. Before describing briefly this original specimen, Cuvier (1804: 300) wrote ‘On n’imagine guère que je sois aussi en état de donner quelques traits de la description du cerveau d’un animal qui paroît détruit depuis tant de siècles: un hasard heureux m’a cependant procuré cette faculté.’ [One can hardly imagine that I am also in a position to give some features of the description of the brain of an animal that appears destroyed for so many centuries: a fortunate happenstance has nonetheless provided me with this faculty]. He also explained straightforwardly that this unusual fossil was formed by the casting of the cranial cavity by the marly sediment that would constitute the matrix.

The task of describing the Montmartre ‘Ornitholithe’ with a natural endocast fell to Gervais (Gervais, 1848–1852a: 230–231; 1848–1852b: pl. 50 fig. 1; 1859a: 410–411; 1859b: pl. 50 fig.1) (Figure 2a), who had previously named Numenius gypsorum (Gervais, 1844) this shorebird likely related to the curlews or godwits (Scolopacidae). However, Gervais (1848–1852a: 230) did not provide any written account of the endocast of Numenius gypsorum. This complete lack of interest in the potential information provided by this part of the specimen is also patent with regard to the other few fossil birds at the disposal of this author that showed, however inadequately, a part of the endocasts. This is the case, for instance, of another Eocene bird from the Paris area (Pantin) that Gervais (Gervais, 1848–1852a: 229; 1848–1852b: pl. 49 fig. 4) named ‘Tringa? Hoffmanni’, which is now known as Ludiortyx hoffmanni and considered close to the gallinules (see in particular Brunet, 1970: 33‐44).

Figure 2.

Historical stages in the imaging and analysis of fossil avian endocasts. (A) Lithography of the first fossil bird (Numenius gypsorum, Eocene, France) in which a natural endocast was noticed (after Gervais 1848–1852b: pl. 50 fig. 1^a; Gervais, 1859b: pl. 50 fig. 1^a). (B) Unpartitioned digital endocast within a skull (Presbyornis pervetus, Eocene, USA) rendered semitransparent (courtesy of WitmerLab; see Zelenitsky et al., 2011: fig. 4). (C) Partitioned digital endocast (Archaeopteryx lithographica, Jurassic, Germany): brain stem, yellow; cerebellum, blue; optic lobes, red; cerebrum, green; olfactory bulbs, orange (courtesy of A. Balanoff; mirrored, see Balanoff et al., 2013: fig. 1)

It seems that the anatomist Gratiolet (1858) was the first to realise the potential of artificial endocasts in the study of mammals, and also birds. Unfortunately, he only published the description of two endocasts from fossil species, both mammals (Gratiolet, 1858; 1859) and, therefore, did not really lead to advancements that were pursued by subsequent authors in the field of avian palaeoneurology. The first depiction in figures of the cast of the endocranial cavity of an extinct bird, in this case a giant moa (Dinornis), fell to Owen (Owen, 1871: pl. 45 figs 11–13; 1879: pl. 91 figs 11–13), who had previously provided observations on the endocranial surface in this and other subfossil birds (Owen, 1848: pl. 52 fig. 7, pl. 53 fig. 7, pl. 56 fig. 11). Only later was the first physical endocast from a non‐avian theropod made (Marsh, 1884: pl. 9 figs 1‐2).

3. STAGNATION

Despite these early works, until the mid‐20th century, a large part of scientists’ attention to avian palaeoneurology was captured by Marsh’s (1880) monograph on Hesperornis, the western bird, and Ichthyornis, the fish‐bird. This is not a result of a lack of specimens of extinct birds that could lend themselves to palaeoneurological study. For instance, some bird cranial remains from the Miocene of the Allier department (France) studied by Milne‐Edwards (1867–1868a; 1867–1868b) (whose father, incidentally, praised the making and examination of endocasts, in mammals at least; Milne‐Edwards, 1868: 204, 211) would have provided a priori excellent material for such research, but they were not put to good use for this purpose. What is more, the existence of natural endocasts of Miocene birds from Baden‐Württemberg and Bavaria were mentioned by Lydekker (1891: 40, 45, 174) but not investigated. As the potential of these and other fossils was not exploited, Marsh’s (1880) study was long held to be a unique source of information about neurosensory capabilities in what were thought, at the time, to be primitive birds.

In this context, it is of little surprise that the analyses of the natural endocast of the London specimen of Archaeopteryx, the Urvögel, by Edinger (1926), de Beer (1954) and Jerison (1968; 1973), which succeeded preliminary observations by Mackie (1863) and Evans (1865; see also Owen, 1863: pl. 1 fig. 1), remained landmark studies with a significant impact on our initial understanding of avian brain evolution. However, they were too taxonomically limited to allow any comprehensive testable picture to emerge.

In the first two‐thirds or so of the 20th century, publications (especially primary sources) on fossil avian endocasts from taxa other than Archaeopteryx, Hesperornis and Ichthyornis were infrequent. In fact, most of the literature in those days consisted of exegeses or mere general considerations about the brain in these taxa only (see Edinger, 1975). The only exceptions included works on much more recent specimens such as Edinger’s (1928; 1942a) study of the palaeoneurology of the bush moa Anomalopteryx and elephant birds Aepyornithidae, Thenius’ (1954: fig. 3) drawing of a natural endocast of a Pleistocene bird (probably a shearwater of the genus Puffinus) from near Athens, Greece, Starck’s (1956) scrutiny of the endocranial morphology in Dinornis, the coastal moa Euryapteryx and other ratites and Stager’s (1964: figs 16, 17) depiction of artificial endocasts of La Brea Tar Pits Accipitrimorphae and one‐dimensional comments about them. Hence, when Edinger (1951) showed that the endocasts described in Marsh’s (1880) seminal work owe more to the imagination than to actual observation, it became clear that our knowledge of the early evolution of the brain in birds amounted to virtually nothing.

4. MODERN RENAISSANCE

Following that relatively stagnated period in the study of the nervous system of fossil birds, a revival of interest in the field has slowly taken place during the latest three decades of the 20th century, during which endocasts from a diversity of birds not only appeared in publications as accessorial objects but also as the focus of attention (e.g. Dechaseaux, 1970; Hoch, 1975; Mlíkovský, 1980; Elżanowski and Galton, 1991; Murray and Megirian, 1998). At the beginning of the 21st century, the first computer graphics reconstructions of extant and extinct bird endocasts were realised (Franzosa, 2004: figs 20, 35–38, 134–135; Domínguez Alonso et al., 2004: fig. 3), a few years after the first digital reconstruction of a non‐avian theropod endocast (Knoll, 1997: pls 12–14; Knoll et al., 1999: fig. 2).

Soon after the advent of this major shift in methodology, the development of digital three‐dimensional (3D) reconstructions boomed, resulting in a fast‐improving understanding of the morphological modifications undergone by the avian brain through time (e.g. Milner and Walsh, 2009; Walsh and Milner, 2011b; Ksepka et al., 2012; Proffitt et al., 2016; Degrange et al., 2018; Figure 2B). This is in keeping with the significant technological and methodological developments that several areas of the field of palaeornithology have been experiencing over the past two decades (Wood and De Pietri, 2015). During this period, not only valuable data on the palaeoneurology of avian crown clades that were not well documented (or not known at all) in this regard, such as Galloanserae (Mlíkovský, 1981,1988; Murray and Megirian, 1998; Milner and Walsh, 2009), Australaves (Mlíkovský, 1980; Degrange et al., 2015) and Accipitrimorphae (Picasso et al., 2009; Scofield and Ashwell, 2009), have been made available, but also our understanding of the brain of stem group taxa (Elżanowski and Galton, 1991; Walsh et al., 2016) has been refined. This improvement of our knowledge also concerns the few taxa that had already been explored palaeoneurologically during the afore‐described stagnated period. Of particular significance are further preparation of the skull of the ‘London specimen’ of Archaeopteryx (Whybrow, 1982) and the subsequent digital restoration of the endocast in this particular fossil essentially (but not only) by Domínguez Alonso et al. (2004), Witmer and Ridgely (2009), Balanoff et al. (2013) and Beyrand et al. (2019).

5. STUDY MATERIALS: BRAIN, SKULL AND ENDOCAST

5.1. Braincase external morphology

Although the external morphology of the braincase is not generally used to describe the endocranial space, in many species of birds, both extinct and extant, the bony wall of the braincase is so thin that its general external shape provides some evidence about the morphology of the endocranial cavity and, therefore, of the brain itself (e.g. Ocampo et al., 2018: fig. 2). This also holds true in some pterosaurs (Edinger, 1941) and some mammals (Schwalbe, 1904). Bühler (1984: 136) noted that, in modern birds, ‘The outer form of the braincase – very much in contrast to the reptilian condition – is relatively similar to the rounded form of the enlarged brain inside’. In fact, the degree of pneumatisation (and consequent thickness) of the neurocranium walls varies considerably among species within extant birds (compare for instance the quail Coturnix coturnix and the zebra finch Taeniopygia guttata in Tahara and Larsson, 2019: figs 3H‐J, R‐V, 5A, B; see also Hesse, 1907: pls 10 and 11; Verheyen, 1953) and is particularly low in some (but not all) small‐sized taxa such as small oscine and suboscine passerines, hummingbirds and swifts (e.g. Chapin, 1949; Winkler, 1979). In Archaeopteryx, the neurocranium wall is very thin (e.g. Rauhut et al., 2018: figs 35C, 37A). The brain ‘inflates’ the braincase in such a way that the morphology of the former can be reconstructed from the form of the latter (Bühler, 1984: fig. 2). Such an observation is also true of enantiornithines, both juveniles (e.g. Knoll et al., 2018: fig. 1), as expected, but also adults (e.g. Zhou et al., 2008: fig. 2A). The brain‐like shape of the neurocranium is so obvious in the putative stem ornithurine Cerebavis (Walsh et al., 2016: fig. 1) that the holotypic braincase was initially interpreted as a ‘brain mould’ (Kurochkin et al., 2006). Yet there are also taxa in which the external surface of the braincase hardly tells anything about the shape and size of the endocranial cavity, especially because of invasive neurocranial pneumatisation. Extinct taxa showing this latter condition include the dodo (Raphus; Owen, 1866: pl. 11 fig. 1; Gold et al., 2016a: fig. 5C), adzebill (Aptornis; Owen, 1848: pl. 52 fig. 7, Owen, 1879: pl. 43 fig. 7) and giant moa (Dinornis; Owen, 1870: pl. 13 fig. 9, 1879: pl. 53 fig. 4, pl. 78 fig. 9; Starck, 1956: fig. 5 and 10; Corfield et al., 2008: fig. 2 C and D; Ashwell and Scofield, 2008: fig. 3f, g and h, 4f, g and h, 5f and 6h), certainly in relation with flightlessness and relaxation of constraints on skull mass. Extant non‐ratite taxa that exhibit extreme amounts of pneumatisation in the skull include Caprimulgiformes and Strigiformes. They generally show extremely thick regions of pneumatised bone around the braincase (e.g. Balanoff, 2004).

5.2. Endocranial cavity

Typically, the palaeoneurology of birds has been studied by examining the endocranial cavity and producing a cast (endocast) or a digital 3D reconstruction of it. The high degree to which the endocranial space, and therefore the endocast, is a faithful reflection of the morphology and volume of the brain in birds (in adults at least) was demonstrated visually by Edinger (1929: fig. 10) in the domestic goose Anser anser domesticus and volumetrically by Iwaniuk and Nelson (2002) in 82 species belonging to various orders, from Struthioniformes to Passeriformes. This has also been stressed more casually by Kappers (1932: 394–396) and many subsequent authors (e.g. Wharton, 2002: fig. 2.3). Elżanowski and Galton (1991:100) remarked, however, that the degree of precision with which the details of the avian brain surface are reflected by its bony covering was not satisfactorily known.

Both physical cross section of the skull (e.g. Toledo da Fonseca et al., 2013: fig. 2E; Karkoura et al., 2015: fig. 6; Tahara and Larsson, 2019: fig. 5A,B) and non‐invasive modern imaging of the brain within the cranium (e.g. Kawabe et al., 2015: fig. 1; Jirak et al., 2015: fig. 2; Li et al., 2018: figs 4 ,5) make it apparent that there is a close match between the sizes and shapes of the brain and those of the endocranial space in individuals of both immature and mature birds (or, in other words, that birds present a brain‐to‐endocranial cavity index – BEC – close to 1; see Balanoff et al., 2016a). Very recently, Watanabe et al. (2019) addressed the reliability of digital endocasts as a proxy for brain size and morphology in archosaurs using a 3D geometric morphometric approach and taking the domestic chicken Gallus gallus as model for birds. They showed that brain and endocast volumes correlate strongly and converge throughout ontogeny: the brain occupies 60% of the endocranial cavity in neonates and over 90% in mature individuals (Watanabe et al., 2019: fig. 3A,B). Rehkämper et al. (1991: 87) hypothesised previously that the absolute brain volume in hummingbirds may be only 4% smaller than the endocranial volume in bats (e.g. Eisenberg and Wilson, 1978: 741). Watanabe et al. (2019) also found that there is a suboptimum correspondence in shape concerning the cerebellum and medulla as opposed to the cerebrum and optic lobes.

Admittedly, structures of the dural venous system interfere to some extent with the intimate relationship between original cerebral morphology and endocast (Balanoff and Bever, 2017: fig. 6). For instance, the cerebellar folia are concealed by the occipital sinus in the endocranial cavity of many, but not all, birds (e.g. Witmer et al., 2008: fig. 6.7A, B; Ksepka et al., 2012: fig. 3A, C‐F, 4A, C‐F, 6A, C‐F; Smith and Clarke, 2012: figs 2A, C, 3A, C, 9A, D, 10A, D, 11A, D; Tambussi et al., 2015: fig. 3B, E, 5A, B, D, 7; Gold et al., 2016a: fig. 3A, B, D, 4A, B, D; Proffitt et al., 2016: 1B; Acosta Hospitaleche et al., 2019: fig. 9B, D; note that this character may be subject to intraspecific variation) (Figure 3). However, the most important features of external brain morphology can be extracted from avian digital endocasts. These include, but are not limited to, the prominence of the Wulst, the size and orientation of the optic lobe and the size of the olfactory bulbs (see below). This is because the mesodermal supporting tissues of the central nervous system, namely the meninges and the venous sinuses within (with the reservation noted above), tend to be very thin in the braincase cavity of birds. As good as the fidelity between endocast and true neural structures is, Walsh and Knoll (2018:61) underlined the necessity of developing a new nomenclature for the descriptions of the former in order to avoid confusion with the terminology used for the latter.

Figure 3.

Partitioned three‐dimensional digital brain endocasts in left lateral (top) and dorsal (bottom) views showing different degrees of faithfulness in the reconstruction of the morphology of the cerebellum. (A) Short‐tailed shearwater (Puffinus tenuirostris). (B) Magellanic penguin (Spheniscus magellanicus). These are related species (Austrodyptornithes), whose endocasts differ mainly with regard to the distinct extent of the occipital dural sinus, which obfuscates the details of the caudal region of the encephalon, thereby obscuring some interesting eco‐ethological information (see Levert, 1973). Brain stem, yellow; cerebellum, blue; labyrinth, purple; optic lobes, pink; optic chiasm and pituitary, red, cerebrum, green; olfactory bulbs, orange

5.3. Fossilisation of the brain itself

We have come to consider of late that actual brain tissues might fossilise under exceptional circumstances (e.g. Pradel et al., 2009). Through spiral computed tomography, Rogers (1998; 1999) suspected the presence of fossilised remains of the actual brain within a natural endocast of the theropod dinosaur Allosaurus. More convincing evidence of fossilised endocranial substance in a dinosaur (an ornithischian, thus a more distant relative of birds) has been put forward recently (Brasier et al., 2017). So, neural tissue of extinct birds may be out there to be found in the right sort of Konservat‐Lagerstätten. Nevertheless, this might be difficult to recognise without the help of complementary analyses, be they microscopic, tomographic or chemical. Kurochkin et al. (2006: 657, 663) suggested that the type specimen of Cerebavis cenomanica was a true fossilised brain, not merely an endocast. The presence of a caudal ‘cavity’ that could be readily interpreted as a foramen magnum and other aspects of this fossil made this interpretation doubtful at first sight and, indeed, it has been subsequently refuted, as the specimen is actually an abraded skull fragment (Walsh et al., 2016).

6. RESEARCH QUESTIONS, MORPHOMETRIC APPROACHES AND NEUROECOLOGY

The field of avian palaeoneurology offers a unique opportunity to explore a number of complex research questions of much interest to palaeontologists and evolutionary biologists alike (Balanoff and Bever, 2017; Walsh and Knoll, 2018). Many neurocranial morphological features could be discussed here, such as those related to the olfactory bulb (Zelenitsky et al., 2011; Corfield et al., 2014) and optic lobe (Torres and Clarke, 2018), the optic (Hall et al., 2009) and trigeminal foramina (Iwaniuk et al., 2009) and the hypophyseal (Edinger, 1942b) and floccular fossae (Walsh et al., 2013), to name but a few. The degree of development of such structures correlates with sensory abilities and other neural functions according to Jerison’s Principle of Proper Mass (Jerison, 1973: 8) and thus is considered to reflect behaviour (see recent review of the Principle of Proper Mass in relation to differences in the sensory capabilities among living birds in Wylie et al., 2015). Investigations of this kind are classic in palaeoneurology.

Here, we will limit ourselves to briefly examining four lines of investigation (distinct but interrelated) within the sphere of interest of avian palaeoneurology that we deem particularly illustrative wide‐ranging topics. These are (1) the forces that shape the vestibular system, (2) the way developmental mechanisms potentially underlie brain evolution, (3) the interplay between the acquisition of flight and, as the case may be, subsequent loss thereof, and the emergence of a large telencephalon and (4) whether and how palaeoneurology can provide information about the process that led to the levels of avian intelligence observed today.

7. VESTIBULAR ECOMORPHOLOGY

The semicircular canals are the part of the vestibular system of the inner ear that provides the central nervous system with sensory input related to the movements of the head. Birds have had a special bearing on our understanding of the function of the semicircular canals, starting with Flourens’ (1825) historical bilateral lesions of the horizontal semicircular canals in pigeons. Bony semicircular canals are sheathed in compact bone within the cancellous bone of the neurocranial walls. Particularly difficult to investigate by the most traditional techniques of palaeoneurology (physical endocast), this aspect of bird neurosensory adaptation has now become more straightforward to explore with the use of modern imaging technology.

Interpreting the arrangement of the semicircular canals in any given taxon is fraught with inherent difficulties, suggesting this morphology may be the result of a compromise between competing selective demands. It has long been hypothesised (e.g. Gray, 1907: 19, 161) that the greater the radius of the semicircular canals, the greater the locomotor agility of its holder. Hadžiselimović and Savković (1964) have provided suggestive evidence for a relationship between morphology and size of the semicircular canals and flight abilities in birds (see also Lewin, 1955). These authors showed that the canals tend to be short and thick with poorly marked ampullae in ‘clumsy flyers’ (e.g. the domestic goose, Anser anser domesticus) and longer and more slender with well‐developed ampullary ends in ‘deft flyers’ (e.g. the stock dove, Columba oenas). As the size of the cerebellum in birds may be related to flight abilities (both airborne and underwater) according to Boire and Baron (1994; see also Walsh et al., 2013), a relationship between cerebellar dimensions and vestibular morphology is conceivable. Nevertheless, the size of the cerebellum cannot be effectively estimated by external measurements of endocasts, notably because the anterior lobe of the cerebellum is entirely hidden by the cerebral hemispheres.

Because the morphology described by the course of the semicircular canals is generally as easily accessible to modern imaging technology in extinct birds as in those species presently living, and because it has the potential to provide valuable data regarding the ecology and behaviour of their holders, it is of high interest to palaeobiologists. Nevertheless, studies such as that conducted by Hadžiselimović and Savković (1964) have merely associated a subjective assessment of the configuration of the semicircular canals with a highly arbitrary characterisation of the flight of the associated species, and there is most likely more than a single factor at play in shaping inner ear morphology.

The lengthening and increased complexity of the course of the semicircular canals throughout the evolution of birds have yet to be demonstrated through 3D quantification of semicircular canals shape and robust statistical procedures to correlate (or not) with flying ability. A recent meticulous analysis of extant avian species by Benson et al. (2017) represents a big step toward that goal. Using μCT scanning and 3D geometric morphometrics, these authors suggested that the labyrinths of living birds are shaped more by spatial constraints in the braincases than by flying style (i.e. the agility of species as flyers or even their capacity to fly at all). Cheung and Ercoline (2018) suggested that the relevant patterns of locomotion that are related to the semicircular canal morphology may not have been captured by Benson et al.’s (2017) methodology. It is not impossible in truth that a pertinent scoring of head movements during locomotion, instead of categorisation scheme of flight behaviour, would largely explain labyrinth morphology in birds. For the time being, the precise reason why there is such a variety of intersections between the three semicircular canals of birds, especially compared with other archosaurs, is not clear yet.

8. ONTOGENETIC ENCEPHALON GROWTH

Examining an enantiornithine embryo, Zhou and Zhang (2004) suggested that the species was precocial (or even superprecocial) on the basis of the presence of well‐developed feather sheets and the large brain. Unfortunately, the specimen was flattened to such an extent that no assertion regarding the endocranial cavity could actually be put forward. In extant birds, different avian hatchling forms have been described, defined by a unique character set documenting not only the growth rate of the skeleton during early life, but also that of the brain (Starck, 1989; 1993). Developmental modes have not influenced exclusively the evolution of brain : body relationships in birds (Iwaniuk and Nelson, 2003); they have had a deep effect on the timing and course of avian brain development, too. Hatchlings with limited development of the brain, in which the optic lobe appears enormous in comparison with the telencephalon, are characteristic of altricial species, whereas a roughly isometric brain growth during post‐hatchling development indicates a precocial species (Portmann and Sutter, 1940; Portmann, 1947; 1962; Sutter, 1943; 1951; Neff, 1972; Bennett and Harvey, 1985; Starck, 1989; 1993; Ricklefs and Starck, 1998).

In the endocasts made from individuals of the greater rhea (Rhea americana) of different ontogenetic age (Picasso et al., 2010), it is evident that the endocast from a chick is largely comparable to those from older juvenile and adult specimens. In this species, which is precocial like other ratites, the main differences are that the telencephalon (forebrain fossa) expands less laterally relative to the mesencephalon and that the Wulst is less prominent in the chick (Picasso et al., 2010: fig. 3). The same is also true of the common ostrich (Struthio camelus) (Romick, 2013: figs 3 and 4). In the domestic fowl (Gallus gallus domesticus), Kawabe et al. (2015) confirmed that the growth rates of each brain region do not change during post‐hatching period, which was expected from the precocial mode of development of this taxon (Figure 4). Given the dearth of information on the evolution of endocast morphology during early development in altricial, semi‐altricial and semi‐precocial species, future research focusing on such species should be encouraged (Beyrand et al., 2019, provided very recently such data about one altricial species, Ficedula albicollis, the collared flycatcher).

Figure 4.

Morphological changes of the brain endocast in post‐hatching period in domestic fowl. The morphology of the endocast changes during the development following hatching according to a pattern similar to the differences observed between species of different adult sizes (with rounded brains in small/young birds, but straighter, somewhat more primitive‐looking ones in large/adult birds). Colour code as in Figure 3

Even though growth series of braincases are hardly ever available for fossil birds, there is a theoretical possibility of discriminating an isolated perinatal braincase of an altricial species from that of a precocial one based on the morphology of the endocast alone. Determining the developmental modes of extinct birds would, in turn, have bearing on determining some aspects of the learning capacity (e.g. vocalisations) of these species because delays in brain maturation typical of altricial taxa probably foster flexible and innovative behaviours (Charvet and Striedter, 2011).

9. PHYLOGENETIC TELENCEPHALON ENLARGEMENT

The brain of Archaeopteryx has long been regarded as intermediate in structure and relative size between the linear and narrow brains of non‐avian extant reptiles and the larger, rounded brains of modern birds (Jerison, 1968; 1973; see also Chatterjee, 1997: fig. 11; Chatterjee and Templin, 2004: fig. 12.3) or in the lower limit of living bird range as far as relative size is concerned (Hopson, 1977; 1980). More data about the encephalisation of non‐flying maniraptoran theropod dinosaurs only acquired during the present century has allowed this view to be debunked, suggesting thereby that the enlarged brain observed in present‐day birds is not the result of a primary adaptation for flight but instead an exaptation (Balanoff et al., 2013; 2016b; 2018; Balanoff and Bever, 2017; Beyrand et al., 2019; a hypothesis formulated previously by Hopson, 1977: 436; see also Witmer and Ridgely, 2007; Milner and Walsh, 2009: 216).

Whole brain size is a blunt instrument when it comes to assessing avian brain evolution. Evidence suggests that the proportions of the telencephalon relative to both the rest of the brain and body as a whole vary considerably from species to species in birds and, indeed, that the volumetric changes of the telencephalon and other main regions of the brain occurred through mosaic evolution in the course of the phylogenetic history of birds (Iwaniuk et al., 2004). Corfield et al. (2012) showed that differential enlargements of telencephalic subregions have taken place across birds during their evolution and that these changes appear to be associated with behavioural, ecological or physiological characteristics. Furthermore, possible ‘trade‐offs’ between brain parts (e.g. Wylie et al., 2015) are not reflected by measuring the whole brain. Therefore, the way the proportions and shapes of individual brain regions and subregions (i.e. the modular subunits that evolve at different rates in different lineages) were altered in the course of bird evolution is more important than the apparent changes of relative dimensions undergone by the brain as a whole. Indeed it is better able to help us discern the factors that triggered the evolution of this organ and the developmental mechanisms that made it possible.

In this context, the recent advances in partitioning digital endocasts into functional neuroanatomical regions (Walsh and Milner, 2011b; Balanoff et al., 2013; 2016b; Walsh et al., 2013; Ferreira‐Cardoso et al., 2017) (Figures 2C and 5) can provide significant insights into the nature and sequences of changes that occurred during the evolution of complex brains in birds. Such investigations may represent a turning point in avian palaeoneurology, a juncture when the field shifts from predominately observational to more hypothesis‐driven and, as such, better able to provide general information regarding the evolution of the brain in these vertebrates. However, this approach is still considerably limited and requires validation by correlation with wet specimen dissections. Investigating how the details of internal brain anatomy may be reflected in the shape of endocasts in birds requires considering major neural pathways. For the time being, preliminary results suggest that components of the tectofugal and thalamofugal visual pathways could be distinguished, as there is a positive relationship between the volumes of the optic tectum and hyperpallium and the surface of the structures by which they are overlain and that are visible on endocasts, the optic lobe and Wulst, respectively (Early and Witmer, 2015; 2016; Early et al., 2019). Studies of brain function, such as that of Gold et al. (2016b) and others, are relevant for understanding the organisation of the modern avian encephalon and unravelling the major changes that took place in the brain of theropod dinosaurs in concert with the emergence of the ability to use powered flight.

Figure 5.

Skull of indeterminate plotopterid (Suliformes) from the upper Oligocene of Kitakyushu, Japan (Kawabe et al., 2014: fig. 1B) in left lateral (top) and dorsal (bottom) views (KMNH VP 200,008; Kitakyushu Museum of Natural History & Human History, Kitakyushu, Japan). Skull rendered semitransparent in order to see the partitioned digital brain endocast within. Colour code as in Figure 3

Using geometric morphometrics, Gold and Watanabe (2018) demonstrated that the loss of flight is not correlated with an appreciable amount of changes in endocast morphology in crown‐group birds, but rather is more constrained by evolutionary history. These authors also suggested that modern birds occupy their own morphospace clearly separated from that of non‐avian theropods and Archaeopteryx (and that the endocast shape of the latter indicates terrestrial habits). Predictably, a more complete sampling of non‐neornithe avialans (beyond Archaeopteryx) might provide a more graded picture of the transition leading to the neornithe‐type of brain from those of basal coelurosaurs. Very recently, Beyrand et al. (2019) recovered a partial overlap in endocranial doming between non‐maniraptoriform dinosaurs, non‐avian Maniraptoriformes and Neornithes, suggesting that the identification of a ‘flight‐ready’ brain is hardly possible.

10. COGNITIVE ABILITIES

It may seem inconceivable to establish a picture of avian palaeoneurology without touching upon the possibilities and limits in assessing intelligence in extinct birds. Despite completely dissimilar brain organisation, birds and mammals share many of the same advanced cognitive abilities (Güntürkün, 2012). Therefore, the origin and evolution of such faculties in birds are of special importance. It has recently been shown that pigeons are as quick as, or even slightly faster than, humans in multitasking (Letzner et al., 2017). These results are in line with others in showing that some bird species rival primates in certain cognitive abilities (e.g. Kabadayi and Osvath, 2017; Von Bayern et al., 2018).

Although this goes against popular belief, the avian performance in complex cognitive tasks is indeed backed up by a number of scientific investigations as well as casual observations. Probably due to endocranial space constraints, the brain of birds generally appears to have more densely packed neurons than that of mammals and a much higher proportion of brain neurons may be allocated to the telencephalon in core landbirds (Olkowicz et al., 2016). How this translates into cognitive abilities is still largely unknown. As information becomes more comprehensive in terms of variations in neuronal density across species, potential correlation between neuron number and cognition should emerge more clearly (Iwaniuk, 2017).

If testing cognition across extant species is not a straightforward procedure, reckoning cognitive capacity in extinct taxa, especially those with no living close relatives, is all the more difficult. By necessity, it often rests on an expression of the relationship between brain and body size, such as the Encephalisation Quotient (EQ; see Jerison, 1973; see also Hopson, 1977; 1980; Hurlburt, 1996; Hurlburt et al., 2013). The EQ of an animal is the ratio of actual brain size to the brain size expected from the body size of this animal. Although EQ is a key metric in palaeoneurology for the analysis of brain size evolution, it is problematic for a range of statistical reasons and, in fact, the very notion that EQ and cognitive abilities relate closely rests upon shaky foundations (e.g. Healy and Rowe, 2007). The empirical data for a link between cognition and brain size in birds is indeed equivocal at present. It is not surprising that such a link is present in some, but not all, studies, given that cognition can be quantified on the basis of a wide range of behaviours and comparative analyses are conducted on different subsets of species.

Based on their relative brain size, birds and mammals have sometimes been lumped together in the same ‘higher vertebrates’ smallest convex polygon (Jerison, 1969: fig. 2, Jerison, 1973: fig. 2.4). This was meaningless due to a number of reasons, in particular the very different neuron size and density in birds and mammals. In fact, EQ comparisons between birds themselves may turn out to be fallacious as well because of the disparate neuronal density observed across the group. For instance, even though the monk parakeet (Myiopsitta monachus) and the rook (Corvus frugilegus) have a similarly high EQ, the former has almost a 30% greater neuronal density in the telencephalon compared with the latter (based on Olkowicz et al., 2016: supporting dataset 1), a priori suggesting dissimilar cognitive skills. Nevertheless, such comparisons within a group of closely related species (e.g. songbirds) may be valid.

Furthermore, calculating well‐founded EQ in living birds is challenging, maybe more so than in any other vertebrate group. To begin with, a number of factors may cause significant variation in body mass in these animals. To cite but three examples, migrant birds usually undergo substantial oscillations in mass over the annual cycle (e.g. Jehl, 1997); the mass of small owls has been shown to be particularly labile depending on the time elapsed since their last meal (Dunning, 1985) and the body mass of growing young can surpass normal adult mass before decreasing prior to fledging in many species of birds (Ricklefs, 1968). At the same time, the measurements of the encephalic mass (or volume; in toto or any subset of it) in birds also comes up against oscillation according to season or other environmental conditions and is influenced by a number of factors such as individual age and experience (e.g. Graber and Graber, 1965; Clayton and Krebs, 1994; Smulders et al., 1995; Healy et al., 1996; Tramontin and Brenowitz, 2000; Roth and Pravosudov, 2009). That all these uncertainties with the acquisition of accurate figures for body and brain masses are exacerbated in fossil birds requires no elaboration.

Nevertheless, recent methodological improvements in the computing of the mass or volume of both the body (allometric: Field et al., 2013; Serrano et al., 2015; volumetric: Brassey and Sellers, 2014; Brassey, 2016) and the brain (Kawabe et al., 2009; 2013b) in extinct species may offer brighter prospects of calculating more exact EQs. For instance, Kawabe et al. (2009; 2013b) demonstrated that brain width significantly correlates with brain volume in many groups, so that the latter can be estimated solely based on simple cross‐sectional metrics. However, this may not hold true in other groups. For instance, some Coraciiformes have exceptionally wide (transversally) but short (rostrocaudally) cerebra, so that the width would not accurately reflect the volume (e.g. Stingelin, 1958: fig. 25).

11. MORPHOMETRIC APPROACHES

Recent generations of digital endocasts of living taxa (e.g. Witmer et al., 2008; Kawabe et al., 2009; 2013a; 2015; Walsh et al., 2013; Picasso et al., 2010; Smith and Clarke, 2012; Romick, 2013; Lautenschlager et al., 2014; Carril et al., 2016; Gaetano et al., 2017; Walsh and Knoll, 2018) have helped considerably to build a reference dataset useful for the examination and interpretation of more ancient avian endocranial morphologies, for instance to test predictions of sensory or locomotor capabilities. In line with another comparative morphological study (Witmer, 1995), paralleling endocast reconstructions of fossil specimens with those of closely related living species (e.g. Ksepka et al., 2012; Stubbs and Ksepka, 2012; Kawabe et al., 2014; Paulina‐Carabajal et al., 2015; Tambussi et al., 2015; Gold et al., 2016a; Torres and Clarke, 2018; Kawabe, 2019), often in an integrative approach, is now a standard framework.

As fossil avian specimens provide only external morphological information on the brain, understanding the brain shape is critical for further discussion about brain evolution in birds. Three‐dimensional geometric morphometrics represent an effective instrument to investigate the differences and similarities between extant avian brain morphologies from which new information about extinct birds may be derived. This is with this end in view that Kawabe et al. (2013a; 2014; 2015; 2017) analysed extant avian brain models using 3D geometric morphometrics. Morphometric studies have their limitations, however. Examining brain morphology independent of what changes might be occurring within the brain is unlikely to reveal much about behaviour, sensory ecology or evolutionary processes. For example, the position and shape of the Wulst of true owls (Strigidae) differs greatly from that of barn owls (Tytonidae), but the Wulst is similar in relative size and appears to support the same aspects of vision in these two families (Iwaniuk and Wylie, 2006: 1,319, tab. 2, fig. 2). Thus, structurally and functionally the Wulst is identical in true owls and barn owls, but the morphometrics of the brains in these groups are vastly different.

Kawabe et al. (2013a; 2015; 2017) suggested that evolutionary change in brain size is a dominant factor influencing the variations of avian brain shape. Kawabe et al. (2013a) also demonstrated that orbital morphology is another factor responsible for shape variation in the avian brain. Thus, brain size and orbital shape are factors that cannot be ignored when assessing encephalic morphology in birds. Not only may the brain affect cranial shape but also functional constraints of the skull can influence brain morphology. Three‐dimensional geometric morphometrics was also used for clarifying the relationship between phylogeny and brain shape in water birds (Kawabe et al., 2014). The width of the cavity that held in life the main mass of the cerebellum and the distance between the tip of one floccular lobe to the other exhibit considerable variation among aquatic birds, and their relative sizes are potentially useful for distinguishing the brains of Procellariiformes and Sphenisciformes from those of Ciconiiformes, Pelecaniformes and Suliformes (Kawabe et al., 2014). Besides, the cerebellar shape changes according to the degree of dependence on aquatic foraging.

Inferring functional organisation of the brains of extinct avian and non‐avian theropods on the basis of what we know of the subdivisions of the pallium in their extant relatives represents a potentially powerful approach yet to be developed in palaeoneurology (see Chen et al., 2013). In this context, results such as those of Fuchs et al. (2014) have the potential to allow moving a step further forward in the exploitation of endocasts to understand the abilities (cognitive and others) and behaviour of ancient birds. Based on a small sample of closely related bird species, these authors suggested that migratory behaviour (as a population‐level trait) correlates with forebrain compositional geometry. Specifically, hyperpallial (Wulst) and hippocampal volumes appear to be larger than predicted from allometric expectations in long‐distance migrants and smaller than expected in birds with modest or no migratory behaviour.

12. NEUROECOLOGY

From the viewpoint of palaeoneurology, most of the avian evolutionary tree is currently undersampled and severely biased in favour of Cenozoic taxa. A fair number of the papers published recently in the area of avian endocranial anatomy have focused on species that became extinct in relatively recent times (geologically speaking), i.e. during the Neogene (see in particular Ksepka et al., 2012; Degrange et al., 2015; 2018; Paulina‐Carabajal et al., 2015).

Some other works even revolved around ‘subfossils’, which offer three‐dimensionally preserved specimens from poorly indurated or friable sediment (or free from matrix altogether), which eases the realisation of both physical and digital endocasts. These studies include species that died out as a result of human activity, such as the moa (Ashwell and Scofield, 2008), Haast’s eagle (Harpagornis; Scofield and Ashwell, 2009), dodo (Gold et al., 2016a) and elephant birds (Torres and Clarke, 2018). Examining the endocasts of these species could aid in interpreting sensory ecology and provide a more accurate reconstruction of their behaviour. The morphology of the endocranial cavity of birds can indeed provide clues related to aspects of ecology in extant species (e.g. Smith and Clarke, 2012).

The brain morphology of endangered species may be relevant to conservation biology (e.g. Corfield et al., 2011). There are many unique and enigmatic taxa for which nothing is known about sensory abilities. Information that could aid in interpreting sensory ecology and providing a more accurate reconstruction of the behaviour of these species could be gleaned from endocasts. For instance, the Plains‐wanderer (Pedionomus torquatus) is a cryptic species endemic to Australia that is rarely observed during the day, the endocast of which may provide ecological data and, therefore, help devise a proactive strategy for protecting the declining population.

13. CONCLUSION

Although avian palaeoneurology has a long history going back to the founding father of vertebrate palaeontology, Cuvier himself, it is only in the last five decades that usable data filling in the almost complete gap in our knowledge of the brains of long extinct birds between Archaeopteryx and Quaternary taxa have been gathered (Dechaseaux, 1970, and later authors). This allowed some light to be shed on the course of avian early cerebral evolution. Nevertheless, as no endocast from common or diverse Mesozoic avian clades such as Confuciusornithidae, Enantiornithes and Hongshanornithidae has yet been published, we still know very little about neurosensory development in early birds. From this point of view, our factual knowledge of the evolutionary history of the avian brain has not significantly progressed since Edinger’s (1951) reassessment of the palaeoneurology of ‘Odontognathae’ about 70 years ago. Recent results on the endocranial cavity of Ichthyornis confirm that the brain was essentially modern‐looking (Field et al., 2018), corroborating that palaeoneurological data from non‐ornithurine avialans are especially needed. This would help test the theory according to which the cerebral enlargement seen in modern birds originated from the acquisition of elevated metabolic rates, itself related to long‐distance powered flight (Shimizu et al., 2017).

Since the turn of the present century the large increase in the use of digital endocasts of modern as well as extinct species of birds has been paving the road for new insights into the evolutionary pathways taken by the avian brain (Franzosa, 2004, and later authors). The widespread use of CT‐scanning sparked a sea change in research on the central nervous system of ancient birds. By allowing the number of publications focusing on the description of endocasts to increase over the last years, this supplementation of the palaeoneurologist’s toolbox is widening the horizons of the field as never before.

Birds have now become a model for analytical approaches developed in palaeoneurology (e.g. Balanoff et al., 2013). The fact that the recently published ‘best practices’ vade mecum in the construction of digital endocasts (Balanoff et al., 2016a) took birds and their close dinosaurian relatives as examples is a milestone that marks that avian palaeoneurology has now reached maturity (see also Balanoff and Bever, 2017). It is hoped that the dynamic trend that avian palaeoneurology has been experiencing over the last decade will continue and that clear methods for how we can use fossil bird endocasts to test predictions about avian brain evolution will eventually materialise. This increased importance of computerised investigations ought not to be to the detriment of fieldwork, though. In fact, as for any other area of palaeontology, the discovery of interesting and suitable fossil specimens (e.g. pre‐Neogene and non‐aquatic birds) may well set the pace of the advancement of our knowledge in this particular field with such a rich history and so vibrant a future.

Knoll F, Kawabe S. Avian palaeoneurology: Reflections on the eve of its 200th anniversary. J. Anat. 2020;236:965–979. 10.1111/joa.13160