Could theropod dinosaurs have evolved to a human level of intelligence?

Noting that some theropod dinosaurs had large brains, large grasping hands, and likely binocular vision, paleontologist Dale Russell suggested that a branch of these dinosaurs might have evolved to a human intelligence level, had dinosaurs not become extinct. I offer reasons why the likely pallial organization in dinosaurs would have made this improbable, based on four assumptions. First, it is assumed that achieving human intelligence requires evolving an equivalent of the about 200 functionally specialized cortical areas characteristic of humans. Second, it is assumed that dinosaurs had an avian nuclear type of pallial organization, in contrast to the mammalian cortical organization. Third, it is assumed that the interactions between the different neuron types making up an information processing unit within pallium are critical to its role in analyzing information. Finally, it is assumed that increasing axonal length between the neuron sets carrying out this operation impairs its efficacy. Based on these assumptions, I present two main reasons why dinosaur pallium might have been unable to add the equivalent of 200 efficiently functioning cortical areas. First, a nuclear pattern of pallial organization would require increasing distances between the neuron groups corresponding to the separate layers of any given mammalian cortical area, as more sets of nuclei equivalent to a cortical area are interposed between the existing sets, increasing axon length and thereby impairing processing efficiency. Second, because of its nuclear organization, dinosaur pallium could not reduce axon length by folding to bring adjacent areas closer together, as occurs in cerebral cortex.

of dinosaur trackways indicate herding behavior and adult care of juvenile members of the species among various theropod, sauropod, and ornithopod dinosaurs (see Figure 1 for an evolutionary tree showing dinosaur classification and relationships) (Hopson, 1977;Ostrom, 1986;Pol et al., 2021). The discovery and study of dinosaur nesting sites has indicated that theropod, sauropod, and ornithopod dinosaur species also nested in groups and engaged in maternal care of hatchling dinosaurs, with some evidence suggesting biparental care (Grellet-Tinner & Fiorelli, 2010;Hopson, 1977;Horner, 1982Horner, , 1987Meng et al., 2004;Pol et al., 2021;Varricchio et al., 2008). Studies of trackways of carnivorous theropod dinosaurs, such as the dromaeosaurid deinonychus, have provided evidence that they hunted in packs (Erickson, 1999;Li et al., 2008). Such social and parental behaviors are largely uncharacteristic of extant reptiles, other than some maternal behavior in crocodilians (Hopson, 1977;Tullberg et al., 2002), and are more reminiscent of the flocking/herding, hunting, and parental behaviors shown by birds and mammals.
Analysis of brain size relative to body size reinforces the conclusion from paleontological behavioral data that at least some dinosaur species rivaled living birds and mammals in behavioral sophistication (Balanoff et al., 2013;Hopson, 1977;Jerison, 1973;Knoll et al., 2021;Ksepa et al., 2020;Russell & Seguin, 1982). For example, although brain size-body size ratios for many dinosaur species tend to be within the range of extant reptiles, many of the theropod dinosaurs show ratios greater than those for living reptiles (for example, allosaurids and tyrannosaurids), and maniraptorid theropod dinosaurs show brainbody ratios that approach those of birds ( Figure 2) (Hopson, 1977;Jerison, 1973;Ksepa et al., 2020). One dinosaur currently known as troodon (Troodon formosus), but once known as Stenonychosaurus inequalis, in particular, shows a brain-body size ratio falling within the lower edge of the largely superimposed avian and mammalian ranges ( Figures 2 and 3) (Hopson, 1977;Jerison, 1973;Russell & Seguin, 1982). Based on the similar brain size-body size scaling exhibited by living basal birds and theropod dinosaurs, Herculano-Houzel (2023) has, in fact, suggested that telencephalic neuron abundance may also scale similarly for these two groups, with larger theropods thus potentially having rivaled some extant nonsimian primate species such as baboons and capuchin monkeys in total telencephalic neuron abundance (Olkowicz et al., 2016). Thus, the notion that dinosaurs were unsophisticated in their behavioral repertoires and that they dominated the land for over 150 million years because evolution had not yet brought forth a better alternative is untenable. The apparently avian level of behavioral and neural sophistication of many dinosaurs, particularly theropod dinosaurs, readily explains their long reign as dominant land vertebrates, until their extinction due to the cataclysmic asteroid strike at the end of Cretaceous period, 65 million years ago (Alvarez, 1987;Alvarez et al., 1980Alvarez et al., , 1984.

COULD A DINOSAUR HAVE EVOLVED TO A HUMAN LEVEL OF INTELLIGENCE?
Dinosaurs were a varied group, and consisted of two major subgroups that were defined by the differences in their pelvic bone structure -the ornithischians and the saurischians (Figure 1). The former included the four-legged armored dinosaurs, such as the ceratopsians, the stegosaurids, and the ankylosaurids, as well as the bipedal duck-billed ornithopod dinosaurs, while saurischians included the four-legged herbivorous sauropods and the two-legged generally carnivorous theropods (Benton, 2004;Sereno, 1998). Although evidence for complex behaviors such as herding and nesting has been found for both ornithischians and saurischians, not all dinosaurs possessed high brain-body size ratios ( Figure 2) (Hopson, 1977;Jerison, 1973).
In particular, theropods tended to show the highest brain-body size ratios among dinosaurs, and likely also the highest telencephalic neuron abundances, as noted above. Some theropods, such as troodon and other troodontid and maniraptoran theropod dinosaurs (e.g., the dromaeosaurids velociraptor and deinonychus), have been documented as having especially large brains for their body size, based on endocast data and skeletal reconstructions, respectively (Figures 2 and 3) (Balanoff et al., 2013;Hopson, 1977;Jerison, 1973;King et al., 2020;Russell & Seguin, 1982;Witmer & Ridgely, 2009). As noted above, troodon was found to conservatively have a brain size-body size ratio at the lower edge of the superimposed avian and mammalian ranges ( Figure 2). Based on eye position and size, troodon and other dromaeosaurid dinosaurs must have also had keen stereoscopic vision (Figure 4a,b) (Hopson, 1977;Russell & Seguin, 1982). Additionally, these dinosaurs had long arms and large hands with fingers that appear to have been capable of grasping objects (Figure 4a,b). This set of traits in troodon and other maniraptorans is reminiscent of the set of traits that are thought to have preadapted simians to undergo selection for increases in the cognitive and motor capacity that allowed more complex manipulations of objects and the planning abilities to facilitate those manipulations (resulting in complex tool making and use), and ultimately the cognitive complexity to allow the syntactically ordered symbolic communication system called language (Dehaene et al., 2022;Herculano-Houzel, 2018;Lovejoy, 1981;Russell & Seguin, 1982;Sherwood et al., 2012;Wilson, 2012). The intellectual gap between humans and all other species is emphasized by even just a few of the uniquely human achievements, such as the harnessing of fire and the invention of the wheel. No other animal species has accomplished these, or anything close. Humans have then used their unique cognitive and linguistic abilities to transmit and accumulate innovation across generations, leading to such exclusively human attainments as agriculture, dwellings, cities, technology, science, mathematics, art, and music (Boyd, 2017). Note that many authors have addressed the use of the term "intelligence" in an ethological context, describing criteria and tests for comparing among species (Hodos, 1988;Reader et al., 2011;Roth, 2015). Why then did troodon or some similar dinosaur not evolve in the direction taken by primates during human evolution? Troodon evolved late in dinosaur evolution, 77 million years ago (Russell & Seguin, 1982), and the catastrophic extinction of dinosaurs 12 million years later may not have allowed sufficient time for such evolution, since it required about 50 million years for modern humans to evolve from early simians (Martin, 1993). But if the catastrophic extinction had not occurred, could troodon have evolved along a simian-like path? Could a human level of intelligence have been reached? At least one paleontologist has F I G U R E 1 Dendrogram showing the evolutionary relationships among the major groups of dinosaurs, and the relationship of dinosaurs to the extinct flying reptiles known as pterosaurs and the still-extant crocodilians. Note the node showing that pterosaurs, crocodilians, and dinosaurs all have a common origin and are collectively referred to as archosaurs. Similarly, all dinosaurs have a common nodal origin, and are divided into two groups based on their hip structure, the bird-hipped ornithischians (red), and the lizard-hipped saurischians (blue). The ornithischians, despite their name, did not give rise to birds, and include several groups, including the armored dinosaurs (stegosaurs and ankylosaurs), the herbivorous ornithopods (which were facultative bipeds and included the iguanodontids, the duckbilled hadrosaurs, and the crested lambeosaurus), the horned ceratopsians, and the dome-headed pachycephalosaurs. Saurischian dinosaurs are divided into two major groups, the long-necked, herbivorous sauropods (which included the massive, four-legged neosauropods such as diplodocus, apatosaurus, and brachiosaurus), and the bipedal, typically (Continues) F I G U R E 1 (Continued) carnivorous theropods. The theropods did give rise to birds, and include various subgroups. Among these, the maniraptorans include small-to medium-sized bipedal dinosaurs with relatively large grasping hands, such as the oviraptorids, the dromaeosaurids (which include velociraptor and deinonychus, notably), and the troodontids. Although archaeopteryx and extant birds were also in the maniraptoran group, their forelimbs obviously became specialized for flight. The dromaeosaurids, the troodontids, archaeopteryx, and extant birds were or are feathered, and make up the paraves.

F I G U R E 2
This image shows a graph redrawn and modified from figure 1 of Hopson (1977), which presents a comparison of brain-body size relationships, expressed on a log-log scale, for living reptiles, birds, and mammals in comparison to dinosaurs, with the range of variation for each group enclosed in a minimum convex polygon. Each polygon is color coded to the named vertebrate group it represents. For purposes of comparisons of interest here, dots are shown for the brain-body ratio for the paravian theropod dinosaur troodon, for the ornithomimid theropod dinosaur ornithomimus, for the large carnivorous theropod dinosaurs allosaurus and tyrannosaurus, for the neosauropod dinosaurs diplodocus and brachiosaurus, for the ornithopod dinosaur iguanodon, for the horned dinosaur triceratops, for the armored stegosaurid dinosaur kentrosaurus, for the extant archosaurian reptile the alligator, and for the mammals, mole, opossum, gorilla, human, and blue whale. Note troodon falls at the lower edge of the overlapping bird-mammal range. Saurischian dinosaurs are indicated by blue dots and ornithischian dinosaurs by red dots. Mammals and non-dinosaur reptiles are shown with black dots. speculated that, in fact, dinosaur humanoids 1 could have evolved from a troodon-type beginning, possibly even reaching a modern human level of intelligence, if given that time (Figure 4b,c) (Russell & Seguin, 1982).
Although other authors have questioned the speculation of Russell and Sequin (1982) based on consideration of the pelvic and hindlimb transformations required to reach the humanoid shape proposed by Russell and Sequin (Naish & Tattersdill, 2021;Paul, 1988), my goal here is to evaluate this speculation from the point of view of the likely brain anatomy of troodon, and dinosaurs in general. Note that consideration of whether dinosaurs could have evolved to a human level of intelligence is of interest for several reasons. First, it addresses the issue of equipotentiality in evolution. Can a similar end point be reached from different starting points? While it is clear that some structures such F I G U R E 3 Image (a) shows endocasts of ostrich, wandering albatross, and troodon from the dorsal view. The troodon endocast shows lengthy olfactory bulbs, differing from the birds in this regard, but otherwise showing a similarly shaped telencephalon and cerebellum. The image is a modified version of figure 3 from Dinosaur Brains by Harry Jerison in the 3rd edition of the Encyclopedia of Neuroscience (Adelman G, Smith BH, eds) (2004), which was provided Dr. Harry Jerison. The image has been modified by removing the background, adding a scale bar, and trimming nonneural fringe from the troodon endocast. Image (b) was also provided by Harry Jerison and shows a perspective view of the troodon endocast. as wings can evolve from different tissues, the starting point nonetheless constrains the outcome (Alberch, 1982;Alexander, 2015;Gerber, 2014;Gould, 2002). It is of interest to know if evolution of intelligence is a trait that is constrained by the features of brain organization in the early members of a line. Second and relatedly, it is of interest to know, or at least consider, if the evolution of the human brain and the cognitive capacity that goes along with it was a unique event that was made possible by features of telencephalic organization incidentally present in ancestral mammals, but not present in the ancestors of dinosaurs and birds. Note also that the issue of whether dinosaurs could have evolved into humanoids is of interest for popular culture reasons, as well, given the many science fiction books, television shows,

F I G U R E 4 Images of Troodon formosus (formerly named
Stenonychosaurus inequalis) and the dinosaur humanoid proposed by Russell and Seguin (1982). Image (a) shows an artist recreation of troodon with a natural habitat as background, while image (b) shows a model of troodon next to a model of the dinosaur humanoid proposed by Russell and Seguin (1982). Note these depictions predate findings indicating that troodon was feathered, but the absence of feathers in these depictions aids in showing the large hand and eye size of troodon. Image (c) shows a drawing of the dinosaur humanoid to more clearly show its human-like features. Note the plantigrade feet in the dinosaur humanoid, as contrasted to the typical theropod/avian digitigrade feet. Image (d) shows the commander of a fictional extraterrestrial race of humanoid reptiles called the Gorn, from the original Star Trek television series, to indicate the intrigue that humanoid reptiles hold in popular culture. Image (a) is by John Sibbick from David Norman: The Illustrated Encyclopedia of Dinosaurs, Salamander Books Limited, New York (1985), and is used with his permission. Image (b) is widespread on the internet, and was originally created by the Canadian Museum of Nature. Image (c) is Dale Russell's dinosaur humanoid as drawn by John Sibbick, from David Norman: The Illustrated Encyclopedia of Dinosaurs, Salamander Books Limited, New York (1985), and is used with his permission. Image (d) is downloaded from the Wikipedia site devoted to the fictional Gorn species, and is widespread on the internet. and movies that include reptilian humanoids as a plot premise, for example, the Gorn commander that did combat with Captain Kirk in the 18th episode of the first season of the original Star Trek television series (the Gorn commander losing, of course) (Figure 4d). Similarly, in the 65th episode of the Star Trek: Voyager television series, the crew of the starship Voyager encountered a sentient reptilian species called the Voth that was revealed to be the descendants of a highly advanced race of ornithopod dinosaurs from earth that had colonized deep space.
While the what-if type of speculation offered by Russell and Sequin (1982) is problematic to evaluate definitively, given that we cannot rerun earth history without the asteroid strike that killed dinosaurs, there are lines of reasoning that can be applied that have as their underpinning a series of credible assumptions, based on existing data or lines of thought. These assumptions are discussed below.

ASSUMPTIONS FOR EVALUATING THE POSSIBILITY OF EVOLUTION OF A DINOSAUR TO A HUMAN LEVEL
Four assumptions are made here in evaluating this issue. These assumptions have to do with the necessary organizational features of a brain if it is to support a human level of intelligence and with the likely organization of the brains of dinosaurs.

About 200 cortical areas as a prerequisite for a human intelligence level
Jerison developed a concept called the encephalization quotient (EQ) by which to use brain-body ratios to assess intelligence (Jerison, 1973).
The EQ essentially measures the extent to which a brain exceeds the size needed to operate the sensory and motor functions of a body of a given size, with that extra brain matter then presumptively related to higher order perception, cognition, and motor planning. In this regard then, the uniquely high intelligence level of humans is associated with an EQ much higher than that of any other species (Jerison, 1973).
Taking a different approach, several research groups have noted that brain size alone can be misleading because neuron density can vary from species to species for a given brain size, with total neuron abundance then being a more likely predictor of computational power than brain size alone (Herculano-Houzel, 2009Kverková et al., 2022;Olkowicz et al., 2016;Sol et al., 2022). In this regard then, neuron abundance in the cerebral cortex (also commonly called neocortex) in humans (16.3 billion) has been reported to greatly exceed that in any other vertebrate group, including gorillas (9 billion) and chimpanzees (6-7 million) (Azevado et al., 2009;Herculano-Houzel, 2009Olkowicz et al., 2016). This is true as well for species with larger brains than in humans but much lower neuron density, such as African elephants with 11 billion by histology-based estimates (Roth & Dicke, 2005) but 5.6 billion by more recent isotropic fractionator-based estimates , and false killer whales with 10.5 billion by histology-based estimates (Roth & Dicke, 2005) but predicted to be more in the 3-4 billion range by more recent isotropic fractionator-based considerations (Kazu et al., 2014). Underlying both the high EQ and the high cortical neuron abundance of humans is the great expansion of the cerebral cortex, not just in total areal extent or weight, but also in the number of separate functionally distinguishable units of which it consists, which are called cortical areas and which are in many but not all cases also cytologically distinct from one another. Cortical areas tend to be modality specific, for example, visual areas, and cortical areas within a given modality tend to be near one another, are organized hierarchically, and engage in increasingly higher order and more abstract information processing (Fellemen & Van Essen, 1991;Van Essen, 2013;Young et al., 1994). Many cortical areas in humans, additionally, seem to integrate information across modalities, and are thus considered associative areas and thought to be a major substrate for higher-order cognition and intelligence (Kaas & Preuss, 2014;Ströckens et al., 2022).
As the first assumption for my argument about the evolution of a dinosaur to a human intelligence level, it is assumed here that a human level of intelligence would require a number of cortical areas, or the equivalent of cortical areas in the case of birds, reptiles, and dinosaurs (as will be discussed below), comparable to that found in humans. An increase in functionally and/or cytologically distinct cortical areas rather than the increase in brain size or neuron abundance per se is here considered to be required to achieve a human level of intelligence (Dicke & Roth, 2016;Rakic, 2009). As noted above, the increase in brain size in humans is assumed, in large part, to be the manifestation of the increase in the number of cortical areas, as well as the increase in the size of the various subcortical areas related to the new cortical areas. This first assumption is based on empirical observations of the abundance of cortical areas in diverse mammalian species and their scoring on various indices of intelligence ( Figure 5a). Basal mammals, such as insectivores, opossums, and monotremes, tend to have a small cerebral cortex that possesses few distinct cortical regions, estimated to be about 20 (Hofman, 2014;Kaas, 2000Kaas, , 2011Kaas, , 2012Kaas, , 2019Kaas, , 2020Kaas & Preuss, 2014;Molnar et al., 2014;Northcutt & Kaas, 1995), and possess low EQs and low cortical neuron abundance (Herculano-Houzel, 2009Jerison, 1973;Olkowicz et al., 2016;Roth & Dicke, 2005). Primates typically tend to have many more cortical areas, higher EQs, and a greater cortical neuron abundance, with the abundance of cortical areas increasing from prosimians to monkeys to great apes to humans (Figure 5a-c) (Herculano-Houzel, 2009Jerison, 1973;Kaas, 2019Kaas, , 2020Kaas & Preuss, 2014;Molnar et al., 2014;Olkowicz et al., 2016;Preuss, 2007). Humans, for example, have cortical areas devoted to language comprehension and production that appear to be lacking in most other primates, but present in rudimentary form in great apes (Sherwood et al., 2003(Sherwood et al., , 2012, and additionally have a greater number of higher order cortical association areas (Glasser et al., 2013;Hofman, 2014;Kaas, 2019Kaas, , 2020Kaas & Preuss, 2014;Van Essen & Dierker, 2007).
The number of distinct cortical areas present in humans is, however, not entirely certain. Brodmann recognized about 50 separate cytologically distinct areas in humans ( Figure 5c) (Brodmann, 1909), but recent multimethod mapping studies in human and higher primates recognize as many as 30 separate visual areas alone (Orban et al., 2004;Van Essen, 2004). Primary and higher order auditory, somatosensory, and motor areas also each appear to be multitudinous in humans, with the added areas performing increasingly higher order information processing and motor planning (Glasser et al., 2016). A most recent estimate based on MRI analysis suggests that humans possess at least 180 separate, functionally distinct cortical areas in a single hemisphere (Glasser F I G U R E 5 Image (a) shows a tree depicting the phylogeny of the cerebral cortex in placental mammals, focusing on the relative size, foliation, and the location of the major cortical areas that are conserved among mammals. Note that as frontal and parietal cortex expands in the line leading to humans, cortical folding increases. The insets at the top depict the hypothetical cortical organization of the mammalian common ancestor (left) (Continues) F I G U R E 5 (Continued) and the primate common ancestor (right). Color coding for cortical areas: Dark blue, primary visual area (V1); light blue, secondary visual area (V2); green, middle temporal (MT) visual area; yellow, primary auditory area (A1); red, primary somatosensory area (S1); orange, secondary somatosensory area (S2). This image is figure 1 in Buckner and Krienen (2013), and is used with publisher permission. Images (b) and (c) show Brodmann's illustrations of the cortical areas in macaque and human, respectively, with humans possessing many more areas than macaque (Brodmann, 1909). Note that the macaque and human brains are not depicted to the same scale. The monkey image is widely available on the internet and widely used by other authors (e.g., figure 2A from Petrides et al., 2012). The human image was downloaded from the internet, where it is widely available (e.g., http://www.systems.neurosci.info/Cortex/brodman). et al., 2016), with there being yet even more possible subdivisions of some of these. Prior histological analyses had reached a similar conclusion (Van Essen, . By contrast, the small amount of neocortex in the prosimian galago is divided into roughly 50 areas (Kaas, 2020), more than twice the number thought to be present in early mammals, and neocortex in marmosets has approximately 100 areas, and that in macaques about 150 areas (Van Essen, Van Essen et al., 2019). It might therefore require evolution of about 200 separate areas for some other lineage to evolve to a human level of intelligence (Kaas, 2020). The reasoning is that increasing the number of cortical areas to 200 (as defined by current methods) appears to be required for the qualitative advance in cortical processing needed to increase intelligence to a human level, with the newer areas beyond those in other primates needed as the substrate to allow higher order information to be abstracted by the new functional cortical areas from the pure sensory input at a human cognitive level. Irrespective of whether future more fine-grained analysis revises the human neocortical area count upward, by phrasing the issue of evolution to a human level in terms of the requisite number of cortical areas (or their equivalents), it makes it possible to evaluate the issue of dinosaur evolution to a human level of intelligence in terms of whether the dinosaur brain design had any built-in features that would have limited this possibility, as discussed in the next section.

Avian-type telencephalic organization in dinosaurs
Second, and importantly for the present argument, it is assumed that dinosaurs must have had an avian type of telencephalic organization.
In this regard, it is important to note, as will be further detailed in a later section, that the cytoarchitectonic organization of the pallium in living birds differs from that in mammals, owing to the divergence of the mammalian lineage and the sauropsid lineage (i.e., reptile and bird) from the stem amniotes about 320 million years ago (Benton & Donoghue, 2007;Dugas-Ford & Ragsdale, 2015;Güntürkün & Bugnyar, 2016;Kemp, 1982;Reiner et al., 2004Reiner et al., , 2005. In brief, the mammalian telencephalon consists of an outer pallial rind, and a rounded more centrally and basally located subpallial region (i.e., the basal ganglia), with the pallium surrounding the basal ganglia dorsally and laterally.
The pallium in mammals includes the typically dorsomedially situated hippocampus, and the more ventrolaterally situated olfactory cortex and amygdala, but the present focus is on the part of mammalian pallium called the cerebral cortex, which makes up the bulk of the pallium that surrounds basal ganglia ( Figure 6). Cerebral cortex has a unique architecture, with its neurons arrayed in six layers that are spanned by the ascending apical dendrites of the pyramidal neurons predominantly making up the layers (Figure 7) (Lund et al., 1979;Shepherd & Rowe, 2017), in contrast to neurons of the striatal part of the basal ganglia just deep to cerebral cortex, which are more uniformly distributed and possess radially symmetrical multipolar dendritic trees confined to the resident nuclear territory of the neuron cell body (Kreitzer, 2009;Wilson & Groves, 1980). The cerebral cortex is also commonly called the neocortex, because it is a new type of cortex only found in mammals, or isocortex due to its relative regional uniformity in cytoarchitecture.
The organization of the avian telencephalon is, however, quite different ( Figure 6). In particular, although a pallium and a basal ganglia can be identified in birds by multiple neurochemical and hodological criteria (Reiner et al., 2004, the neurons of the pallium are not laminar in their perikaryal distribution or columnar in their dendritic orientation, but rather are relatively uniformly distributed, possess radially symmetric multipolar dendritic trees ( Figure 7) (Ahumada- Galleguillos et al., 2015;DeVoogd & Nottebohm, 1981;Durstewitz et al., 1999;Fortune & Margoliash, 1992;Reiner et al., 2001;Tombol et al., 1988;Watanabe et al., 1983), and are organized into many separate nuclei of specific connectivity and function (Reiner et al., 2004. As a result, the pallium and basal ganglia are both nuclear in their organization in birds, a feature that misled earlier comparative neuroanatomists to conclude that the telencephalon in birds was largely a hypertrophied basal ganglia, as has been discussed at length elsewhere (Jarvis et al., 2005;Reiner, 2005Reiner, , 2009Reiner et al., 2004Reiner et al., , 2005. In addition to a small dorsomedial hippocampal region and a small ventrolateral amygdaloid and olfactory region, the nuclear pallium in birds consists of two major territories, a dorsomedial region called the Wulst and a more ventrolateral region called the dorsal ventricular ridge (DVR), the latter of which in birds consists of two major subregions called the mesopallium and nidopallium that together jut into the lateral ventricle and largely obscure it ( Figure 6) (Dugas-Ford & Ragsdale, 2015;Reiner et al., 2004Reiner et al., , 2005. The nidopallium is separated from the underlying striatum by a thin cell-free glial lamina, as is the mesopallium from the underlying nidopallium, and the Wulst from the underlying mesopallium. 2 Thus, the pallium in birds consists of cytoarchitectonically distinct, separate slabs of tissue that are stacked upon one another, with the multipolar dendrites of neurons in each slab confined to that slab, except to some extent at their adjoining borders ( Figure 7). Within each slab are separate nuclear groups of varying distinctness. This organization differs from that of mammalian F I G U R E 6 Images (a) and (b) show schematic line drawings of mid-telencephalic transverse brain sections of rat and pigeon, respectively. Image (a) shows the basic features of mammalian telencephalic organization and image (b) shows the organization of avian telencephalon. In each schematic, the speckled region represents pallium, the striped region represents striatum, and the checked region represents globus pallidus. The pallium in mammals largely consists of neocortex. The pallium in birds at the level shown mainly consists of the Wulst (hyperpallium) and dorsal ventricular ridge (DVR) (mesopallium and nidopallium). These images are figure 1b,d from Reiner et al. (2004). Hp, hippocampus. Images (c-f) present a series of low-power images of transverse sections through the telencephalon of adult rat (c, e) and adult pigeon (d, f). Images (c) and (d) of Nissl-stained sections for rat and pigeon show that the bird telencephalon conspicuously lacks a laminated pallial structure resembling the mammalian neocortex along its outer surface. Instead, the pallial part of the telencephalon of birds resembles the striatal part of the mammalian basal ganglia in its histological appearance. Image (e) shows that the striatum in mammals is characterized by enrichment in dopaminergic terminals, as immunolabeled for tyrosine hydroxylase (TH). Image (f) shows that only the basal part of the telencephalon in birds immunolabels intensely for TH and is therefore the only part of the telencephalon in birds that is striatal in nature. Scale bar = 2 mm in (b) (applies to c-f). Images (c-f) are figure 2A-D from Reiner et al. (2004).

F I G U R E 7
Illustration showing the dendritic morphology of the main neuron types of cerebral cortex (a, b), the fiber organization of cerebral cortex (c), and dendritic morphology of the main neuron type of avian pallium (d, e). Image (a) shows a schematic of pyramidal neurons and their ascending apical dendrites traversing overlying layers, as redrawn from Purves et al. (2008). Image (b) shows a Golgi-labeled pyramidal neuron of layer 5 and a Golgi-labeled stellate neuron of layer 4 from cat, using figures 5A and 2A, respectively from Lund et al. (1979). Image (c) shows a schematic of fiber labeling in cerebral cortex, using an image widely available in the internet. Images (d) and (e) show the location and projections (d) and dendritic morphology (e) of a biocytin-filled IT-type pallial neuron in pigeon from figure 5 of Reiner et al. (2001). This morphology is typical of pallial projection neurons, as also noted in publications cited in the text, meaning that the dendrites of individual pallial projection neurons, except potentially at the borders between regions, are confined to the resident territory of the perikaryon.
cerebral cortex, the functional equivalent of Wulst and DVR, which consists of a single slab (i.e., the neocortex) that surrounds the basal ganglia, and that is organized into layers of like-type neurons whose dendrites cross layers, most commonly the radially disposed dendrites of the pyramidal neurons of layers 2/3 and 5/6 ( Figure 7). As will be detailed in a latter section, despite these differences between neocortex and avian pallium in their cytoarchitectonic configurations, they do resemble each other in showing a radial organization in fiber networks that span and interconnect across layers in mammals ( Figure 7) and across pallial regions/slabs in birds (Stacho et al., 2020).
It should also be added that, like in mammals, pallial projection neurons are excitatory and use glutamate as their neurotransmitter, while basal ganglia projection neurons are inhibitory and use GABA as their neurotransmitter (Reiner, 2002;Reiner et al., 2004).
Note that the pallium in birds also possesses a cytoarchitectonically distinct caudobasal region termed the arcopallium, which for presentation purposes will be treated as part of the DVR, as others have done (Dugas-Ford & Ragsdale, 2015;Iwaniuk & Hurd, 2005). The telencephalon in all living reptiles, as well, possesses a DVR, but one that tends to be smaller than in birds (Bruce, 2007;Reiner, 1991Reiner, , 1993. Extant reptiles also possess a pallial region comparable to Wulst, but unlike in birds it lies above a pronounced lateral ventricle and it is thus relatively more distinctly differentiated from DVR than is Wulst in birds (Bruce, 2007;Reiner, 1991Reiner, , 1993Medina & Reiner, 2000).
Moreover, the reptilian Wulst equivalent is laminated and the apical dendrites of its neurons, which resemble pyramidal neurons of mammalian neocortex in their morphology, traverse its layers radially. Thus, it is cortical in structure, and accordingly, it is called the dorsal cortex (Bruce, 2007;Reiner, 1991Reiner, , 1993. Because all members of living reptile groups and all birds possess a DVR, and because dinosaurs were reptiles and theropod dinosaurs were the forerunners of birds (Chen et al., 1998;Chiappe, 1995;Xu et al., 2003), it is an inescapable conclusion that a DVR is a defining feature of the reptilian-avian clade (i.e., sauropsids), including dinosaurs ( Figure 8). Consistent with this, the brain shape in troodon endocasts and that of other similar theropods such as velociraptor resembles that in birds, with the paired telencephala viewed from above having a characteristic triangular shape, as compared to the more elongate shape of the paired mammalian telencephala ( Figure 3) (Hopson, 1977;Russell & Seguin, 1982). DVR cytoarchitectonic organization varies among sauropsids, however, from a simple laminar organization in the primitive lizard-like reptile Sphenodon to a more complex cellular organization in crocodilians and birds (Northcutt, 1978(Northcutt, , 1981Reiner & Northcutt, 2000). As birds are now regarded as descended from small maniraptoran theropod dinosaurs, the same group to which troodon belonged, maniraptoran theropod dinosaurs (if not all dinosaurs) must have also possessed a complex cellular DVR within the telencephalon, and a Wulst or dorsal cortex as well. Note that a Wulst is not unambiguously evident in the troodon endocast ( Figure 3), but nonetheless may be present in some incipient form not evident from the gross shape of the dorsal endocast surface. Consistent with this inferred bird-like pallial organization, maniraptoran dinosaurs seem to have shown an avian level of behavioral sophistication (Erickson, 1999;Hopson, 1977;Horner, 1982Horner, , 1987Meng et al., 2004;Pol et al., 2021;Varricchio et al., 2008). Moreover, telencephalic regional organization in the only living member of the archosaur reptilian group to which dinosaurs belong, namely crocodilians, shows a very high similarity to that in birds, based on molecular markers, immunomarkers, and connectivity, with crocodilian DVR showing clear evidence of containing a mesopallium and nidopallium (Brauth, 1984;Pritz, 1975Pritz, , 1980Pritz & Northcutt, 1977). Thus, it is highly likely that troodon and related theropods possessed a DVR with a mesopallium and nidopallium that was divided into a large num-F I G U R E 8 Cladogram illustrating the evolutionary relationships among mammals, birds, and reptiles, focusing on the relationships among birds, living reptiles, and dinosaurs. Note that birds and dinosaurs are phylogenetically grouped among reptiles. Given that the cladogram branching pattern for birds, extant reptiles, and dinosaurs shows that a dorsal ventricular ridge (DVR) must have been already present in the common ancestor of birds, living reptiles, and dinosaurs, it is inescapable that the dinosaur telencephalon must also have possessed a DVR. It is more difficult to know whether dinosaurs also possessed a Wulst. Archeopteryx but troodontid endocasts provide suggestive evidence that early avians but not the immediate theropod forerunners of birds such as troodon possessed a Wulst (Jerison, 2004;Balanoff et al., 2013). Thus, troodon may have possessed a structure more resembling reptilian dorsal cortex than avian Wulst, or been intermediate. Note that contemporary analysis indicates turtles to be a sister group of archosaurs (Rieppel, 1999;Chiari et al., 2012;Crawford et al., 2012;Wang et al., 2013).
ber of separate cell groups specialized for various sensory and motor functions, as well as a Wulst (or dorsal cortex) that was devoted to additional sensory and motor functions. Note that Briscoe and Ragsdale (2018) did not detect unambiguous molecular evidence of an arcopallium in alligator, nor is there hodological evidence. Nonetheless, given the close phyletic relationship of birds and maniraptoran dinosaurs, and their behavioral similarities, it seems likely that an arcopallium had evolved in this theropod dinosaur group, if not earlier. Finally, the basal ganglia in dinosaurs was also likely to have been at an avian grade of organization (Bruce et al., 2016;Reiner, 2002;Reiner et al., 1998), but the basal ganglia is not especially germane to the arguments presented here, and so it will not be considered further.

Importance of information processing by the different neuron types in a cortical column
Third, it is assumed that the interactions between neurons within different layers of neocortex are critical to its function. The neocortex is a specialized set of six cytoarchitectonically distinct layers, with each layer containing functionally and hodologically distinct neuron types (except layer 1, which is largely cell free), with the entire ensemble designed to perform a set of computational operations that enable information processing and motor planning (Bennett, 2020;Hanganu-Opatz et al., 2021;Harris & Mrsic-Flogel, 2013;Larkum et al., 2018). Irrespective of the precise algorithms for these operations, the processing power of neocortex is typically assumed to derive from a major feature of its organization-the interactions of the functionally distinct neuronal elements within columns that span the depth of neocortex (Harris & Mrsic-Flogel, 2013;Larkum et al., 2018). The neuronal elements in a column form the fundamental processing units of neocortex, with layer 4 neurons receiving thalamic input and typically possessing a multipolar dendritic shape, with layer 2/3 neurons with pyramidal neuron dendritic morphology receiving input from the thalamorecipient layer 4 neurons, with the layer 2/3 neurons projecting to the layer 5/6 neurons with pyramidal neuron dendritic morphology, and with the layer 5/6 neurons typically projecting out of the neocortex (Figures 7 and 9) (Da Costa & Martin, 2010;Kaneko et al., 2000;Mountcastle, 1997;Rockland, 2010;Shepherd, 2013). The cortical columns are arrayed next to one another, and each cortical area consists of many such columns. For any given cortical area, each column is devoted to a particular part of sensory, motor, or cognitive space for the functional modality of that area. Note that columns are real morphological entities in some demonstrable sense for some sensory systems (such as primary visual cortex in primates or the barrel fields of rodent somatosensory cortex) (Hubel & Wiesel, 1968;Woolsey & van der Loos, 1970), but to a large extent they are theoretical constructs used to characterize the processing mechanisms of neocortex (Da Costa & Martin, 2010;Rockland, 2010), with no firm boundaries between adjacent columns.
It is important to note that each neocortical layer interfaces seamlessly with its adjacent layers and there are no glial barriers separating layers, with dendrites of any given layer extending into other layers, and the ascending apical dendrites of the pyramidal neurons in layer 2/3 and 5/6 traversing the layers above them (Figure 7). The neocortex is, thus, a single cytoarchitectonic entity. In any case, because the neocortex is organized into sets of neurons layered parallel to the pial surface, the functionally distinct neurons of neocortex within each column are placed at close distance to one another-within 1-3 mm, depending upon the mammalian species and cortical region (Amunts et al., 2012;Balaram & Kaas, 2014;Dicke & Roth, 2016;Hofman, 1989;Laughlin & Sejnowski, 2003;Preuss et al., 1999). This and the entry of the dendrites of layer 5/6 neurons into the territory of layer 2/3 neurons shorten axonal conduction times between the neuronal elements making up a column, and keep them relatively invariant across mammalian species, with human neocortex being thickest among primates at about 3.5 mm. The short conductions times presumably benefit the speed, efficiency, and effectiveness of the processing carried out within a cortical column. Note that an increase in the number of columns making up a single cortical area would increase the grain with which information from a periphery of constant size could be analyzed. This might improve sensory discrimination, but it would not yield a qualitative change in information processing. Increasing the number of cortical areas is presumed to be required to yield the qualitative change needed to increase "intelligence," in which higher order information is abstracted by the new functional cortical areas from the pure sen-sory input. Note that each layer of neocortex for each of the major basic cortical areas has a correspondent in a particular nuclear cell group in the avian pallium, as discussed in more detail in a later section ( Figure 9), and presumably this was true of the dinosaur pallium as well (Dugas-Ford & Ragsdale, 2015;Reiner, 2000;Reiner et al., 2005).

Long axon lengths hinder information processing
Finally, it is assumed that increasing axonal length between neurons carrying out a critical processing operation is harmful to the effective execution of this operation, due to increasing delays in transmission because of greater axon length (Cherniak, 1995;Dicke & Roth, 2016;Hofman, 2014). This assumption, which has empirical and theoretical support, is partly the basis of the view that the close spacing of the disparate neurons within cerebral cortex furthers its ability to process information (Hofman, 2014). This rule, however, is thought to apply for all neural circuits. Note, however, that the proximity of the cortical layer neuron types to one another does not invariably favor mammalian cortical over avian nuclear pallial organization as a substrate for intelligence, since many large-brained, pallial neuron-rich avian species such as crows and parrots show complex problem-solving ability that rivals that of monkeys and/or apes, and clearly exceeds that of many mammals, including, for example, rodents, rabbits, insectivores, and opossums (Auersperg & von Bayern, 2019;Balakhonov & Rose, 2017;Bobrowicz et al., 2021;Cnotka et al., 2008;Emery, 2006;Emery & Clayton, 2004a, 2004bHerculano-Houzel, 2017;Kabadayi & Osvath, 2017;Olkowicz et al., 2016;Pika et al., 2020;Ströckens et al., 2022).
The studies of Pepperberg (1999) with African grey parrots, in particular, show they can meaningfully use human words, and grasp simple numerical and relational concepts. Similarly, New Caledonian crows can make tools out of twigs to retrieve inaccessible food, and pass this skill to other crows (Hunt, 2000;Weir et al., 2002), and ravens show domain-flexible future planning (Kabadayi & Osvath, 2017) and socially instructed learning (Pika et al., 2020).
Still, there may be some differences in cognitive ability between these avian groups and the more cognitively capable mammalian species. For example, mirror-guided self-recognition is a sophisticated cognitive trait seen consistently in humans, even as early as 18 months of age, that is thought to indicate an awareness of self, but only in a limited number of other mammalian species (Emery & Clayton, 2004b;Ristau, 2013). Among nonhuman mammals, it has been consistently seen in individual chimpanzees, orangutans, and dolphins, and in Asian elephants, gorillas, and gibbons mirror-guided self-recognition was seen in some studies and individuals but not others (Emery & Clayton, 2004b;Prior et al., 2008;Ristau, 2013). By contrast, among avian species thought to parallel apes in their intelligence, common ravens, azure-winged magpies, carrion crows, jungle crows, and African grey parrots have not shown mirror-guided self-recognition (Emery & Clayton, 2004b;Vanhooland et al., 2022;Wang et al., 2020), although such self-recognition was reported in Indian crows (Parishar et al., 2021) and in three of five Eurasian magpies (a corvid) tested (Prior et al.,

F I G U R E 9
Images and schematics illustrating the location and connectivity of hodologically similar neuron types in primary auditory cortex (A1) of adult mouse and the primary pallial thalamorecipient auditory area of the adult chicken nidopallium (Field L). Images (a) and (b) show the locations of A1 and Field L, respectively. Image (b) is image 49 from the coronal series of sections from Brainmaps.org. Image (c) shows that auditory thalamic input to A1 ends on layer 4 thalamorecipient neurons (color coded green and shown as multipolar), which project to layers 2/3 intracortically projecting neurons (color coded purple and depicted as pyramidal), which themselves project to extracortically projecting neurons of layers 5/6 (color coded red and depicted as pyramidal), which project to striatum, thalamus, and brainstem. Note, by the scale bar, that the axonal pathways between these cortical neuron types are less than a millimeter. Image (d) shows that auditory thalamic input to Field L ends in layer 4-type thalamorecipient neurons of subfield L2 (color coded green and shown as multipolar), which project to nearby layers 2/3-type "intracortically" projecting neurons of subfields L1 and L3 (color coded purple and shown as multipolar), which themselves project to another set of layers 2/3-type "intracortically" projecting neurons (color coded purple and shown as multipolar) located at the lateral edge of the caudal nidopallium, which then project to layers 5/6-type "extracortically" projecting neurons of medial arcopallium (color coded red and shown as multipolar), which project to striatum, thalamus, and brainstem. These illustrations show, by the scale bars, that the axonal pathways between these cortical neuron types in chicken are substantially more than a millimeter. Note that image (d) is not meant to be a comprehensive diagram of Field L circuitry, but rather is intended to show how some of the key layer-equivalent elements are located in different pallial regions at some distance from one another. A more comprehensive Field L circuit diagram would also include mesopallial elements (see text), but these are not included for sake of simplicity. Including mesopallial elements to the circuit, however, would reinforce the point that layer-equivalent elements are located in different pallial regions at some distance from one another in the avian pallial design. The schematic plan for (c) and (d) is adapted from figure 3 of Dugas-Ford and Ragsdale (2015). 2008), but not in another study of a larger magpie cohort (Soler et al., 2020). Moreover, although crows mimic chimpanzees in tool use, some authors have shown that chimpanzees can be taught to communicate using human sign language (Gardner & Gardner, 1980), but there is no such evidence for language use by corvids, suggesting that while corvids may cognitively match chimpanzees in some regards, they do not do so in all regards. Of course, at least one African gray parrot has shown ability to communicate using human words (Pepperberg, 1999), but as noted above they have not shown mirror-guided selfrecognition. Thus, while the proximity of the cortical layer neuron types to one another (and thus the short axons distances between neuron types within a layer) may not invariably favor mammalian cortical over avian nuclear pallial organization as a substrate for intelligence in living birds, as will be discussed in more detail in later section, the axon distance problem may become more constraining as more cortical areas (and more pallial neurons) are added in the avian nuclear design.

LINES OF REASONING-CONSTRAINTS ON CORTICAL EXPANSION
Based on these four assumptions, I argue that the brain of troodon or any other dinosaur would have been limited in its ability to add the number of efficiently functioning cortical areas (or their equivalent in a nuclear pallium) requisite to achieve a human level of intelligence.
This conclusion is based on three lines of argument. These arguments are based on considerations of increased axonal lengths that would have been required by a dinosaur (i.e., avian) type of pallial organization, on the inability of a dinosaur telencephalon to fold to achieve reduction in the length of axon paths between adjacent sets of neurons equivalent to those in a given layer of adjacent cortical areas in a mammal, and because the 3D nuclear plan to the architecture of the dinosaur pallium may not have readily lent itself to a precise point-topoint topographic mapping between regions, as possible in the case of the sheet-like cerebral cortex in mammals, which possibly would hinder the efficiency of information processing. The case for each of these three major conclusions is detailed below.

Avian pallial organization versus mammalian cortical organization
Before detailing the reasoning behind the conclusion that axonal length considerations argue against the notion that troodon would have been able to give rise to a lineage that evolved to a human level of "intelligence," it is necessary to review the major features of pallial organization in birds and contrast them to those in mammals in more detail than above, because pallial organization in theropods such as troodon was highly likely to be very similar to that in extant birds. In mammals, the cerebral cortex is the pallial structure that covers most of the outer surface of the telencephalon ( Figure 6). As noted, the cerebral cortex consists of a series of functionally distinct layers of neurons layered on top of one another within a depth of 1-3 mm of one another, depending on cortical region and species (Figures 6, 7, and 9) (Dicke & Roth, 2016;Karten & Shimizu, 1989). Within each layer, neurons are more or less functionally alike, while across layers they are functionally distinct (Dicke & Roth, 2016;Dugas-Ford & Ragsdale, 2015;Güntürkün & Bugnyar, 2016;Reiner et al., 2004Reiner et al., , 2005. One of the manifestations of this within-layer functional uniqueness is laminar specialization in receiving or transferring information to other brain areas (Figure 9a,c).
Layer 4 of neocortex (also called the granule cell layer because of the typically small size of its neuronal perikarya) in all parts of neocortex receives the sensory or sensory-motor information from the dorsal thalamus, or other neocortical areas (Figures 7 and 9a,c)  project to basal ganglia and other neocortical areas, but not outside the telencephalon, while those in lower layer 5 do project outside of telencephalon (Reiner et al., 2003;Shepherd, 2013). The layer 5a neurons are commonly referred to as intratelencephalically projecting neurons (IT-type) because of their connectivity, while those of 5b have been termed pyramidal tract neurons (PT-type) because they give rise to the pyramidal tract (Reiner et al., 2003).
The organization of the avian equivalent of cerebral cortex is different in many important regards, and cladistic considerations suggest that it seems likely that the equivalent of cerebral cortex in dinosaurs shared these avian features (Figure 7). In birds, the equivalent of the cerebral cortex comprises two major structures, the Wulst and DVR, as noted above ( Figure 6) (Dugas-Ford & Ragsdale, 2015;Güntürkün & Bugnyar, 2016;Reiner et al., 2004Reiner et al., , 2005. The Wulst makes up about 20% of the cerebral cortex equivalent, and the DVR makes up the remainder, though this varies among avian groups (Hamaide et al., 2018;Iwaniuk & Hurd, 2005;Iwaniuk et al., 2008;Sayol et al., 2016). While the Wulst has been referred to as quasi-laminar, it is important to note that the neurons of Wulst do not possess the architecture of cerebral cortex neurons in terms of the dendrites spanning across the Wulst quasi-lamina Dugas-Ford & Ragsdale, 2015;Medina & Reiner, 2000;Watanabe et al., 1983).
Moreover, some Wulst divisions are separated by neuron-free glial zones, unlike the layers of cerebral cortex (Reiner et al., 2004). Thus, both Wulst and DVR consist of numerous separate territories that are nuclear in nature, with the main territories for Wulst being the apical hyperpallium (HA), the intercalated nucleus of the apical hyperpallium (IHA), and the densocellulare hyperpallial nucleus (HD), and for DVR being the nidopallium, mesopallium, and arcopallium Dugas-Ford & Ragsdale, 2015;Güntürkün & Bugnyar, 2016;Reiner et al., 2004Reiner et al., , 2005. Within Wulst, the IHA is the main recipient of sensory thalamic input, the HD gives rise to outputs to other parts of Wulst or DVR, and the HA gives rise to outputs to DVR, basal ganglia, and/or brain stem from Wulst Dugas-Ford & Ragsdale, 2015;Güntürkün & Bugnyar, 2016;Karten et al., 1973;Nakamori et al., 2010;Reiner & Karten, 1983;Reiner et al., 2001Reiner et al., , 2005Shimizu et al., 1995;Veenman et al., 1995). Both HA and HD receive input from IHA, and HD projects to HA as well. In the case of DVR, some subdivisions within the nidopallium receive sensory thalamic input, other parts of the nidopallium and all of mesopallium give rise to input to other parts of DVR, and arcopallium gives rise to outputs to basal ganglia and/or brain stem Dugas-Ford & Ragsdale, 2015;Reiner et al., 2005).
As detailed below, thalamorecipient nidopallium projects to nearby non-thalamorecipient nidopallium. Thus, the avian equivalent of mammalian cerebral cortex contains the three major neuron types found in cerebral cortex, namely the thalamorecipient, the intracortically projecting, and the extracortically projecting types Dugas-Ford & Ragsdale, 2015;Güntürkün et al., 2021;Reiner et al., 2001Reiner et al., , 2005Shanahan et al., 2013), but they are not arrayed in sheets of like-type neurons layered at specific depths from the pial surface within a single architectonic entity, as in the case of cerebral cortex (Dugas-Ford & Ragsdale, 2015;Reiner et al., 2004Reiner et al., , 2005. Focusing then on what this means for the information processing unit of neocortex called the cortical column, the neuronal cell groups in birds that together make up the equivalent of the set of neurons in a cortical column for a given area of neocortex are found in separate cell clusters (or nuclei), which in some cases are juxtaposed but in most cases are distant from one another (Figures 7 and 9d).
To make this clearer, it is useful to go into more detail on specific sensory circuits. Each major sensory thalamic input to the avian DVR ends within a specific cell group within the nidopallium (Figures 9b,d and 10).
For example, visual input from nucleus rotundus of the thalamus (which itself receives tectal input) ends in the so-called entopallium within the nidopallial region of the DVR ( Figure 10) (Fernández, Ahumada-Galleguillos, et al., 2019;Hunt & Künzle, 1976;Karten & Hodos, 1970;Karten & Revzin, 1966;Revzin & Karten, 1967), while auditory input from nucleus ovoidalis of the thalamus (which itself receives input from the inferior colliculus) ends within L2 of the so-called Field L of the nidopallium (Figures 9d and 10) (Karten, 1967(Karten, , 1968Wild et al., 1993). Each of these thalamorecipient cell groups projects to a shell of neurons that surround the thalamorecipient cell group  Wild et al., 1993). In the case of Field L, Calabrese and Woolley (2015) have confirmed that the hierarchical organization of the auditory receptive field density and sound selectivity in Field L2 relative to Fields L1/3 resembled that for layers 4 and 2/3, respectively, in A1 auditory cortex. Among the targets of the perientopallial belt and L1/L3 are within-modality IT-type neurons at the dorsolateral margin of the pallium (Figures 9d and 11b,c). These regions in turn project to striatum (Reiner et al., 2001;Veenman et al., 1995;Wild et al., 1993), as well as to specific within-modality parts of the arcopallium (Figures 9d and 11b,c) (Dugas-Ford & Ragsdale, 2015;Fernández, Morales, et al., 2019;Wild et al., 1993). Note that the neurons at the dorsolateral margin of the pallium that project to striatum and arcopallium may correspond to layer 5a IT-type pyramidal neurons in mammals that only have projections within telencephalon Reiner et al., 2003).
Bolstering this possibility is the finding that both the IT-type neurons of layer 5a and the PT-type neurons of layer 5b in mammals receive layer 2/3 input, with the IT-type projecting to the PT-type, in addition to striatum (Shepherd, 2013). Note that the neurons at the dorsolateral margin of the avian pallium (a territory in which regional distinctions among pallial zones can be uncertain) might also correspond to a class of layer 2/3 neurons in mammals that receive input from other layer 2/3 neurons, but for the purposes of developing the reasoning in this article, I will consider them to be like layer 5a IT-type neurons in mammals.
In any case, the visual and auditory regions within the arcopallium receiving dorsolateral pallial input project to brainstem and basal ganglia (Fernández, Morales, et al., 2019;Zeier & Karten, 1971), and thus seem akin to layer 5b/6 neurons of mammalian cerebral cortex (Reiner et al., 2003(Reiner et al., , 2010. This interpretation of arcopallial neuron identity has been confirmed using molecular markers of layer 5b/6 neurons Dugas-Ford et al., 2012;Dugas-Ford & Ragsdale, 2015), as has the identification of the avian pallial territories resembling thalamorecipient, cortical layer 2/3, and layer 5a neurons. The pallial components of the entopallial circuit appear in particular comparable to a visual area in postrhinal temporal cortex in mouse (Beltramo & Scanziani, 2019), while those of Field L are comparable to primary auditory cortex (Calabrese & Woolley, 2015), based on the similarities between mammals and birds in the thalamic inputs (Yamamoto & Reiner, 2007). Of note, the cell groups that are akin to the different layers of postrhinal temporal cortex and to the different layers of primary auditory cortex are typically separated by many millimeters in the avian telencephalon, but in mammals the comparable neurons are much nearer to one another since they reside in cortical columns (Figures 9   and 11b,c). The distributed feature of the visual and auditory nuclei of the pallium in birds that collectively correspond to specific cortical visual and auditory areas in mammals also is the case for the input of trigeminal somatosensory information into the avian telencephalon (Figures 10a and 11a) (Fernández et al., 2021;Wild & Farabaugh, 1996;Zeigler & Karten, 1973), with separate interconnected pallial cell groups akin to those of the different cortical layers in mammals.
Thus, considering the avian pallial circuit for a given modality to consist of one thalamorecipient area, one layer 2/3-like area (with possibly at least one more in mesopallium, as will be discussed in the next para- area that may correspond to layer 5a (or an additional 2/3-like area), and an area with layer 5b/6 neuron types, to derive the number of interconnected neuronal sets equivalent to a cortical area in avian F I G U R E 1 0 Panels showing schematic diagrams of frontal views through the cerebrum, diencephalon, midbrain, and/or hindbrain of the pigeon depicting the cell groups that make up the ascending components of the major trigeminal somatosensory circuit to the DVR, the tectofugal visual pathway to the DVR, and the major auditory circuit to the DVR. Bas, nucleus basalis; BG, basal ganglia; Ento, entopallium; EW, nucleus of Edinger-Westphal; FLM, fasciculus longitudinalis medialis; Hp, hippocampus; Hy, hypothalamus; IC, inferior colliculus; LP, lateral pontine nucleus; M, mesopallium; M5, motor nucleus of the trigeminus; MP, medial pontine nucleus; N, nidopallium; Ov, nucleus ovoidalis; Pr5, principal nucleus of the trigeminus; Rt, nucleus rotundus; SN, substantia nigra. F I G U R E 1 1 Schematics depicting the higher order pallial components of the trigeminal somatosensory (a), tectofugal visual (b), and primary auditory (c) inputs to the DVR, shown in sagittal view with anterior to the left. The schematic for each sensory input illustrates the projection of the thalamorecipient layer 4-like neuronal population (color coded green) to the layers 2/3-like belt surrounding the thalamorecipient population (color coded purple), which in turn projects to the layer 5a-like external nidopallium (also color coded purple), with the external nidopallium then projecting to the layers 5b/6-like arcopallium (color coded red). Note that the L1/3 Fields are shown as closer to their target nidopallial field than they actually are (for sake of schematic convenience), and their actual proximity is better illustrated in Figure 8d. The schematic illustration of these (Continues) F I G U R E 1 1 (Continued) sensory circuits makes the point that the cortical layer equivalents in the avian pallium tend to be at some distance from one another. Note that although the arcopallium is the output region of the DVR, HA (not shown in this illustration) is the output region of Wulst. A, arcopallium; BG, basal ganglia; Cb, cerebellum DVR, dorsal ventricular ridge; HA, hyperpallium apicale; OT, optic tract; Ov, nucleus ovoidalis; Pr5, principal nucleus of the trigeminus; Rt, nucleus rotundus. pallium, one needs to divide the total number of functional areas in avian pallium by at least 5. Thus, taking the Shanahan et al. (2013) estimate of 48 separate pallial regions in pigeons, this would represent the equivalent of no more than 10 cortical areas in a mammal, putting pigeons in the range of primitive living mammals such as insectivores.
The above description of DVR circuitry highlights circuits within nidopallium and connecting to arcopallium, but the mesopallium is also a major DVR territory that is interconnected with the nidopallium and arcopallium. For example, the thalamorecipient entopallium projects radially to overlying mesopallium, which in turn projects radially back to entopallium and its nidopallial belt (Ahumada-Galleguillos et al., 2015;Fernández, Ahumada-Galleguillos, et al., 2019;Kreutzfeld & Wild, 2005). Auditory nidopallium and trigeminal nidopallium are similarly connected radially to overlying mesopallium (Atoji & Wild, 2011;Fernández et al., 2021;Stacho et al., 2020). Mesopallium, in turn, projects to dorsolateral nidopallium and arcopallium (Atoji & Wild, 2009, 2011Wild & Farabaugh, 1996). Mesopallium also receives input from Wulst (Atoji & Wild, 2012;Shimizu et al., 1995), and thus appears to integrate information from DVR and Wulst and operate as a pallial association territory. This interpretation is consistent with evidence from molecular markers and by its connectivity that it principally contains IT-type neurons (Atoji & Karim, 2014;, and from the evidence that its relative size seems to correlate with indicators of cognitive complexity in birds (Mehlhorn et al., 2010;Ströckens et al., 2022;Timmermans et al., 2000). It is uncertain how to draw parallels between mammalian cortical circuitry and the interconnections of mesopallium with nidopallium and arcopallium, but my goal here is neither to exhaustively detail avian pallial circuitry nor to identify one-to-one correspondents in mammals for all aspects of avian pallial circuitry. Regardless of whether the IT-type neurons of mesopallium are compared to layer 2/3 or layer 5a IT-type cortical neurons, the distances in birds between mesopallium and its input sources and its output targets are greater than for either of these two possible mammalian correspondents. Because this point can be made without the added complexity of including mesopallial circuitry in Figure 9d, that particular illustration focuses only on nidopallial circuitry. Note that the Shanahan et al. (2013) estimate of 48 separate pallial regions in pigeons includes mesopallial areas. Since mesopallium, by the above considerations, should be regarded as containing part of the circuit equivalent to a cortical column, this would mean that at least five separate pallial regions in birds contain the correspondent of a cortical column in mammals. Thus, the 48 separate pallial regions in pigeon of Shanahan et al. (2013) would represent the equivalent of 10 or fewer cortical areas in a mammal, depending on how many mesopallial areas are to be regarded as part of the equivalent of a cortical layer constituent for a given cortical area.
Thus, the neuronal elements making up the functional equivalent of a cortical column are typically spread more than 2 mm apart in the cerebral cortex-equivalent pallium of living birds (Wulst and DVR both), and in the case of some they are separated by even more. For example, in Figure 9d showing chicken DVR, the layer 2/3 neurons surrounding Field L2 are about 4.5 mm distant from the layer 5a-like neurons of dorsolateral pallium, which in turn are about 2 mm distant from the layer 5b/6-like neurons of arcopallium. By contrast, the distances between comparable cell types in mouse A1 (Figure 9c) are less than half a millimeter. The above-noted distances in birds increase severalfold in more large-brained birds such as carrion crow and jungle crow, based on images shown in brain atlases for these species (Izawa & Watanabe, 2007;Kersten et al., 2021). Similarly, the Wulst in barn owl is 5-to 6-mm thick (Pettigrew, 1978), and images of barn owl brain available online at www.brainmaps.org show the distance between the center point of the thalamorecipient layer 4-like neurons of IHA and the center point of the layer 2/3-like neurons of HD to be about a millimeter, and about twice that for the distance from HD center point to the center point of the layer 5/6-like neurons of HA. By contrast, in mammals the distances for comparable cell types is much less to begin with in the neocortex of mammals with small-sized to moderately sized brains (a fraction of a millimeter), and does not increase greatly in larger brained species, since neocortex does not thicken substantially from small-brained to large-brained mammals (Amunts et al., 2012;Balaram & Kaas, 2014;Dicke & Roth, 2016;Hofman, 1989;Laughlin & Sejnowski, 2003;Preuss et al., 1999).
At this point in my presentation, it would be useful to comment on the issue of the mammalian homologues of the avian Wulst and DVR, an issue that has been debated for about 100 years (Reiner et al., 2004. The primary visual region, primary somatosensory region, and primary motor region in Wulst in birds and in dorsal cortex in reptiles are widely regarded as homologous to the comparable territories in the dorsal-most sector of neocortex in mammals (Butler et al., 2011;Dugas-Ford & Ragsdale, 2015;Medina & Reiner, 2000), but the mammalian homologues of the visual and auditory territories in the DVR are not firmly established, with some suggesting they reside in temporal neocortex (Butler, 1994;Butler et al., 2011;Karten, 1991Karten, , 1997Karten, , 2015Reiner, 1993Reiner, , 2000Reiner et al., 2005;Shimizu & Karten, 1990;Yamamoto & Reiner, 2007) and others suggesting they reside in the claustroamygdaloid territory (Abellán et al., 2009;Bruce & Neary, 1995;Puelles, 2022;Puelles et al., 2000Puelles et al., , 2016Striedter, 1997). For purposes of the present consideration of the evolution of a dinosaur humanoid, the mammalian homologues of Wulst/dorsal cortex and DVR are not truly germane, since the relevant issue is their functional equivalence, and it is widely agreed that the Wulst and DVR function like cerebral cortex (Calabrese & Woolley, 2015;Dugas-Ford & Ragsdale, 2015;Herculano-Houzel, 2020;Reiner et al., 2005;Stacho et al., 2020;Wang et al., 2010).
Finally, on the topic of avian pallial organization, Stacho et al. (2020) recently reported using high-resolution polarized light microscopy that the primary sensory areas in DVR possess a radial, column-like organization in the connections between the thalamorecipient zones and their surrounding shells, and across the thalamorecipient zones, layer 2/3-like and layer 5/6-like territories in the case of Wulst (Figure 12).
A similar point has previously been made for DVR visual, auditory, and trigeminal circuitry and Wulst visual and somatosensory circuitry based on connectivity studies and recording studies by other authors (Ahumada-Galleguillos et al., 2015;Atoji et al., 2018aAtoji et al., , 2018bCalabrese & Woolley, 2015;Fernández, Ahumada-Galleguillos, et al., 2019;Fernández et al., 2021;Reiner & Karten, 1983;Wang et al., 2010). The common presence of radial organization in avian pallium and mammalian cerebral cortex has been used to explain why avian pallium can support a level of behavioral sophistication on a par with that supported by mammalian cerebral cortex (Herculano-Houzel, 2020; Rose, 2022;Stacho et al., 2020). It is important to note, however, that the radial pallial organization in birds differs in four important respects from the radial organization of mammalian cerebral cortex. First, the radial organization of DVR in birds does not include the layer 5a-like neurons of dorsolateral pallium or the layer 5b/6-like arcopallium of birds, while layers 5/6 neurons are included as a radial component of the cortical column in mammals (Figures 7c and 12). Second, the radial organization of the avian thalamorecipient territories and their belts in DVR differs from the organization of the thalamorecipient layer 4 of mammalian cerebral cortex and the layer 2/3 cortical neurons to which they project, because the dendrites of layer 4 and layer 2/3 cortical neurons extend into each other's territories, but this is not the case for DVR thalamorecipient neurons in birds in relation to their layer 2/3-like belt neurons, except at borders between regions (Figure 7c).

F I G U R E 1 2 Schematic illustrations
showing rat brain (a) and a pigeon brain (b) in side view, depicting the radial pallial organization described in Stacho et al. (2020). Image (a) of rat brain shows the radial organization of six-layered mammalian neocortex, with thalamorecipient layer 4 color coded purple and the supragranular layers 2/3 and the infragranular layers 5/6 color coded blue. The dark blue strands passing through the cortical layers indicate the radial, columnar organization of its processes (dendrites and axons), while the white strands orthogonal to the dark blue strands indicate the horizontal connections among cortical columns. Image (b) shows the Wulst and DVR of the avian pallium. The primary sensory input layer 4-like thalamorecipient zone of the Wulst, namely the intercalated nucleus of the hyperpallium apicale (IHA), is color coded in purple, as are the primary sensory input layer 4-like thalamorecipient zones of the DVR, namely the trigeminal sensory nucleus basalis, the tectofugal visual entopallium, and the auditory Field L2. Note that the thalamorecipient zones of the DVR reside in the deeper part of the nidopallium, and each is surrounded by the layer 2/3-like neurons of their nidopallial belts. The layer 2/3-like neurons of mesopallium reside above the nidopallium. The dark blue strands passing through the Wulst pseudolayers and from thalamorecipient nidopallium through nidopallium belt and into mesopallium indicate the radial, columnar organization of the axonal connections among these regions. As for mammals, the white strands orthogonal to the dark blue strands indicate the horizontal connections among pallial regions. Note that, although not indicated in image (b), the layer 5a-like neurons of the nidopallium caudale (NC) receive input from nidopallium belt and mesopallium, with the NC then projecting to the layer 5b/6-like neurons of arcopallium. The inputs to NC and arcopallium, which complete the neocortex-like circuitry of the DVR, are not radially organized. Moreover, nonsensory parts of DVR are also not radially organized. Thus, the radial organization of avian pallium only partly mimics that of mammalian neocortex. Cb, cerebellum; DVR, dorsal ventricular ridge; HA, hyperpallium apicale of the Wulst; HD, hyperpallium densocellulare of the Wulst; IHA, intercalated nucleus of HA; NC, caudal nidopallium. The schematic is adapted from figure 8C of Stacho et al (2020), and is used with permission of author and publisher.
Note that although Stacho et al. (2020) refer to this organization across nidopallium and mesopallium as columnar, the organization differs from what I have defined as columnar in the present article. Moreover, as we had already noted, the Wulst organization for its visual and somatosensory zones is radial, but again it is not columnar as that term is used for mammalian neocortex and as I have used it here, and in the case of Wulst as well the dendrites of the multipolar neuron types in each pseudolayer do not extend into the territory of another, except potentially at borders between regions (Watanabe et al., 1983).
The columnar organization of neocortex, as well as its compactness, yields short axonal distances for communication between neuron types (Figures 7c and 9c). Third, despite the finding that the radial organization of the connectivity for thalamorecipient and layer 2/3-like neuronal populations in bird DVR resembles that for mammalian neocortex, the thalamorecipient and layer 2/3-like neuronal populations remain more distant from one another in birds than is the case for mammalian neocortex (Figure 9). This is also true for the pseudolayers of the Wulst, especially in larger brained birds such as owls and corvids (Izawa & Watanabe, 2007;Kersten et al., 2021). Fourth, Stacho et al. point out that the radial organization of DVR and Wulst only pertains to the aforementioned sensory circuits. It does not pertain to nonsensory associative circuits. By contrast, associative cortical areas in mammals have radial columnar organization just like sensory and motor cortical areas do, and it has been suggested that associative areas are especially critical for intelligence (Kaas & Preuss, 2014;Ströckens et al., 2022).
So, while it is valuable to note the similarity between neocortex on one hand and DVR/Wulst on the other in radial organization of connectivity between layers or their equivalents, and how this may help explain the remarkable behavioral abilities of birds, there are still noteworthy organizational differences, and the crux of my study has to do with these differences.

Axon length consideration and the cortical column equivalent in sauropsids
Thus, separate nuclear cell groups in the avian Wulst and DVR that are interconnected and carry out a single modality-specific function represent the equivalent of a cortical area. For example, Figure 11b shows that the entopallium, its belt, the dorsolateral belt target, and part of the arcopallium (four separate pallial territories in bird) are the equivalent of a single cortical area, specifically, postrhinal temporal visual cortex in mouse (Beltramo & Scanziani, 2019), and the individual sets of neurons in each that are connected across the separate nuclear groups are the equivalent of cortical columns. Importantly, as noted in the prior section, the neurons making up the equivalents of a cortical column in birds are on the whole more distant from one another than the neuronal elements in a mammalian cortical column (Figures 9   and 11). Clearly, however, the avian pallium, even with its greater axonal length between the layer-equivalent populations of a specific cortical area equivalent, can subserve sophisticated behaviors such as flocking, tool making, vocalization, and parental care (Auersperg & von Bayern, 2019;Balakhonov & Rose, 2017;Emery, 2006;Emery & Clay-ton, 2004a, 2004bPepperberg, 1999;Pika et al., 2020). It may be that the very high neuronal density in the avian pallium serves to reduce the axonal distances between layer-type groupings that would otherwise occur (Herculano-Houzel, 2017;Olkowicz et al., 2016). It also may be that the surprising switch from use of VGLUT1 as in mammals and non-archosaur reptiles to VGLUT2 as the vesicular transporter used by glutamatergic pallial neurons to load glutamate into synaptic vesicles may have facilitated synaptic transmission between interconnected parts of a "cortical" column in avian pallium, and thereby compensated for the synaptic delays caused by axon lengthening associated with the avian-type pallial organization Fremeau et al., 2004;Islam & Atoji, 2008;Sarkar & Atoji, 2018;Zhang et al., 2018).
Moreover, it may that the mesopallium is an avian innovation that serves as an association area to subserve complex cognition. It may also be that the distances between the layers of neocortex, or their equivalents in bird pallium, are not as critical to the functioning of the cortical column (or its equivalent in birds) as widely supposed, at least at the distances exhibited in extant avians.
Living birds, however, have only a limited number of interconnected pallial regions that form the equivalent of cortical areas. Although the precise number is unknown, in birds with simpler pallia, such as pigeons or chickens, it is likely to be comparable to that in a rodent-insectivore grade of mammal, based on behavioral considerations, and this is consistent with the number of separate pallial areas reported by Shanahan et al. (2013) for pigeons, as noted above. An avian pattern of telencephalic organization in dinosaurs could have proven a limiting factor in the ability of troodon descendants to acquire the neural substrate needed for language and complex planning and tool use (i.e., attaining a human intelligence level), due to the information processing inefficiencies introduced by adding additional areas to an avian-type pallium, causing ever-increasing axonal length between them. Increases in the number of pallial areas in either a mammalian pallium or an avian-type pallium are thought to occur by parcellation from existing areas, with the new pallial areas receiving input from new thalamic nuclei parcellated from existing thalamic sensory nuclei and/or from the preexisting pallial sensory areas (Ebbesson, 1980;Rakic, 2009). This seems to be the case for cortical areas that have been added during the evolution of various mammalian lineages from a more basal condition (Kaas, 2011(Kaas, , 2019Kaas & Preuss, 2014). In birds, the individual layer equivalents of any such new areas would have to be spatially separated from one another, since this is characteristic of an avian-type pallial organization. The various cell groups that would have to be added to achieve the equivalent of adding new cortical areas within an aviantype pallial plan would need to be interpolated among existing areas, presumably near the same layer equivalent-type neurons of the same modality ( Figure 13). Note, however, that this would require increasing distances between neurons corresponding to the various layer equivalents of preexisting areas. This would be especially true in the case of projections from the sensory belt regions (layer 2/3 equivalents) to the layer 5a equivalents at the dorsolateral telencephalic surface and from this latter population to the layer 5b/6 equivalents in the arcopallium ( Figure 13) (Veenman et al., 1995). This increase in axon length between the cell fields making up the scattered cell types constituting F I G U R E 1 3 Panel (a) shows a schematic summary diagram in sagittal view of the higher order pallial components of the trigeminal input (T) to the dorsal ventricular ridge (DVR), the tectofugal visual input (V) to the DVR, and the auditory input (L) to the DVR, in each case with the thalamorecipient layer 4-like territory color coded green, the nearby layers 2/3-type territories and the more remote layer 5a-like territories of the external nidopallium color coded purple, and the layer 5b/6-like arcopallial regions color coded red. Arrows indicate the interconnected regions. Panel (b) shows the consequences of a hypothetical pallial expansion, with four new sets of interconnected pallial territories added. In this instance as well, the input recipient territories (either from thalamus or another pallial area) are color coded green, the layer 2/3 and 5a territories are color coded purple, and the output territories of the arcopallium are color coded red. Arrows are shown interconnecting the input, layer 2/3 and 5a territories and arcopallial output territories, with arrows color coded by the type of neuronal population from which they originate. Note that with the interpolation of new areas in panel (b), path lengths between preexisting interconnected regions are increased. The problem would be yet more exacerbated with the addition of even more new regions. As for Figure 11, I should note that although the arcopallium is the output region of the DVR, HA (not shown in this illustration) is the output region of Wulst. DVR, dorsal ventricular ridge; HA, hyperpallium apicale. the equivalent of a single cortical area would become especially problematic as the number of cortical area equivalents approached a human ∼200 cortical areas (with its attendant 16.3 billion neurons).

Measurements of the approximate interlaminar distances in a variety
of mammalian and avian species with known neocortical/pallial neuron abundances or known numbers of cortical areas (or their avian equivalents) would be needed to scale the magnitude of the distance increases with increasing abundance of neocortical/pallial neurons or cortical areas approaching a human level. Nonetheless, as a rough approximation, one can conjecture that to go from the presumptive 10 cortical area equivalents in a chicken to a human 200 (and thus at least a 20fold larger pallium), the distance between the Field L belt layer 2/3-like neurons to the dorsolateral pallial layer 5a-like neurons would roughly increase to about 90 mm, compared to the approximately 1.5 mm seen in human A1 (Hackett et al., 2001). With dinosaur humanoid body size comparable to that in a human, this distance would likely increase yet more, because larger bodies yield larger brains to support and manage the sensory and motor needs of the body. Similarly, the distance between the dorsolateral pallial layer 5a-like neurons to the layer 5b/6like neurons of arcopallium would become about 40 mm in an avian pallial design with a human-matching 200 cortical areas, compared to the approximately 0.2 mm distance seen between the midpoints of upper and lower layer 5 in human A1 (Hackett et al., 2001). The greater size and axon length requirement than with a mammalian type of cortical organization might also have introduced problems of metabolic and thermoregulatory support of such an expanded brain (Heldstab et al., 2022;Hofman, 2014;Karbowski, 2014;Laughlin & Sejnowski, 2003;Pontzer et al., 2016;Sherwood et al., 2012), and thereby additionally limited the extent of expansion possible. Note that although an increase in pallial neuron density in a primate-like theropod lineage could theoretically have compensated for the problems described above, there is likely to be a limit on neuron cell body packing density, due to the need for space for blood vessels for metabolic support (Kaas & Preuss, 2014;Kumar et al., 2022;Ventura-Antunes & Herculano-Houzel, 2023), dendrites for the plasticity underlying learning and cognition (Chklovskii et al., 2002;Harris, 1999;Segal, 2017;Spires-Jones & Knafo, 2012), and glia for neuronal and axonal support (Hughes, 2012;Lenz & Nelson, 2018;Newman, 2003;Sierra et al., 2014;Simons & Nave, 2016;Wilson, 1997).
Finally, it should be noted that the argument presented above posits that the anatomy of the DVR would lead to axon length problems as cortical areas were added to the DVR. It could be argued that the Wulst might not be plagued by such problems because it is quasi-laminar.
It might be further argued that evolution in a primate-like theropod might have favored increasing the number of areas in the Wulst rather than the DVR, and thereby mitigated the axon length problem attending DVR expansion. While this is certainly true, it remains the case that dinosaurs must have already possessed a substantial DVR that they could not readily transform to being more Wulst like or do away with. Thus, while preferential expansion of the Wulst might have been a more efficient path (one that is chosen in owls at least for vision and somatosensory systems, but not in other birds such as ravens, crows, and parrots) (Iwaniuk & Hurd, 2005), the very presence of a DVR limits the efficient expansion possible because it would not be trivial to "undo" the DVR. This is especially the case when it is considered that much addition of areas has occurred in the mammalian functional correspondent of DVR, namely the temporal neocortex (Butler, 1994;Butler et al., 2011;Karten, 1997Karten, , 2015Reiner, 1993Reiner, , 2000Shimizu & Karten, 1990;Yamamoto & Reiner, 2007). Moreover, the Wulst is only quasi-laminar, and thus beset by the same axon length problems as in the case of the DVR, though to a considerably lesser degree.

Folding consideration
Cortical areas in mammals are interconnected with one another in a hierarchical fashion within modality and across modalities (Fellemen & Van Essen, 1991;Van Essen, 2013;Wang & Clandinin, 2016;Young et al., 1994). This interaction seems to be one of the underpinnings of the higher order processing ability of primates and humans. Increasing axon length requirements, however, would certainly be a potential problem in the case of the addition of new cortical areas, since they would need to be interconnected, at increasingly remote distances, and this could constrain the benefits of such interareal connections.
Consistent with the premise that an increase in cortical areas leads to a disproportionate increase in interareal cortical connections, cortical white matter volume increases more greatly with brain size increase than does cortical gray matter volume in primates, and in humans represents about 50% of overall cortical volume (Frahm et al., 1982;Hofman, 2014;Kaas & Preuss, 2014;Van Essen et al., 2019). One major way in which the architecture of cerebral cortex as a surface structure has made it possible to reduce this difficulty, to a large extent, is by its ability to fold into gyri and sulci (Wang & Clandinin, 2016), as typically occurs in species with a large cortical surface area, with only a few exceptions such as manatees (Welker, 1990). Van Essen has shown, in fact, that two adjacent cortical areas that are interconnected typically are often on the opposite sides of a gyrus (Van Essen, 1997). By becoming apposed at their deep surfaces, such cortical areas greatly reduce the length of axon required to interconnect them ( Figure 14).
In general, such folding allows an increase in the cortical surface area and the number of cortical areas relative to the rest of the brain, without a net gain in the length of axon required to interconnect any set of related adjacent cortical areas, thereby combatting the axon length problem otherwise created by an increase in cortical areas (Bayly et al., 2014;Cherniak et al., 2004;Hofman, 2014 In the case of mammalian cerebral cortex, only adjacent areas can fold to reduce the distance between them. Thus, with an increase in cortical areas, the problem of connecting increasingly distant areas, and thereby increasing axon length, is not entirely eliminated. A way in which this problem seems to have been addressed is for local sets of cortical areas to be highly interconnected (referred to as a small world network), with more sparse connections between small world sets of cortical areas (Wang & Clandinin, 2016). In this light, it is of interest that avian pallial areas seem to use the same strategy in their functioning, since Shanahan et al. (2013) report that pigeon pallium is organized into small-world networks with a connective core of hub nodes. Thus, avian pallium uses at least one of the strategies for reducing axon length and volume burden with increasing pallial volume that is also used in the case of mammalian neocortex. It seems likely that any primate-like evolution of theropod dinosaurs might also have employed this partial solution to the problem of increasing axon length between area equivalents with increased number of cortical area equivalents.
Note also that the nuclear configuration of the avian pallium may itself be an adaptation that serves to reduce distances between pallial regions corresponding to layer equivalents of different neocortical areas, distances that would be greater in the case of a more linear arrangement of side-by-side pallial regions ( Figure 15). It may be that this feature of a nuclear pallial plan for reducing the distances for horizontal connectivity in the absence of folding, together with the small world network plan and the high neuron densities, helps explain why the avian pallial design, despite its issues with axon length between cortical layer equivalents, is able to support the sophisticated behaviors seen in crows and parrots.

Mapping of 2D space consideration
Many types of sensory space, such as visual and somatosensory, lend themselves to being mapped in a 2D X-Y layout. This makes 2D sheets of neurons such as characterize mammalian cerebral cortex ideal for mapping and processing information of this type (Kaas, 1997). Orderly projections between cortical areas are also facilitated by the 2D architecture of mammalian cerebral cortex, since this makes topographically ordered fine detail point-to-point transmission of information possible (Kaas, 1997). This is even true if the information transmitted does not involve a map of sensory space. The sheet-like architecture of mammalian cerebral cortex may further its ability to efficiently process information because of precise point-to-point topographic mapping considerations or because of a greater efficiency of lateral inhibition with 2D neuroarchitecture (Biswas et al., 2022;Helmstaedter et al., 2009;Kaas, 1997;Knudsen et al., 1987;Wilimzig et al., 2012). By contrast, the 3D architecture of the nuclear DVR and to a lesser extent the Wulst may pose problems for mapping of sensory space, the efficiency of lateral inhibition, and/or the fine-grained point-to-point projections between different pallial regions (Atoji et al., 2018a(Atoji et al., , 2018bManger et al., 2002;Pettigrew & Konishi, 1976;Reiner & Karten, 1983;Shimizu et al., 1995;Stacho et al., 2020;Wild, 1987). Possible examples reflecting this design constraint of avian pallium are that (1) receptive fields in owl visual Wulst are no smaller than 2 • , while in mammals they can be as small as 1 • in V1 (Clark & Columbo, 2020;Smith et al., 2001), and (2) the size of the iso-orientation domains in barn owl visual Wulst is larger than those found in area V1 of cat and monkey (Liu & Pettigrew, 2003). Further studies that detail topographic precision in avian pallial circuits throughout their key steps are needed to address these F I G U R E 1 5 Schematic illustration showing that in a nuclear organization such as in the avian pallium, adjacent interconnected areas representing the same point in sensory space are brought about 30% closer together than they would be in a cortical design with no folding. This feature may reduce axon length in the nuclear design of the avian pallium for connections between regions, but not totally overcome the limitations of the inability to fold and thereby bring adjacent areas into juxtaposition, and the limitations of the great distances between the dispersed correspondents of neuron types found in different cortical layers in mammals.
possibilities (Hellmann & Güntürkün, 2001;Fredes et al., 2010;Witkovsky et al., 1973). At least one sensory area of DVR in birds, the auditory area (Field L), has evolved to be a flat sheet within the DVR, Fitting so many flat sheets within a 3D space would require an Escherlike degree of tessellation that does not seem simple to evolve or to control developmentally. While the failure to consistently map information on a 2D space has not been a critical problem for living avians, this factor could have imposed a further limit on the efficacy of information processing as the pallium expanded and been an additional factor that constrained the ability of troodon, or a similar dinosaur, to evolve to a human cognitive state.

CONCLUSIONS AND APPROACHES FOR FALSIFYING CURRENT CONCLUSION
In summary, then, I have offered several lines of reasoning for the conclusion that neither troodon nor any other dinosaur could have begun a primate-like lineage that evolved to a human level of intelligence.
The argument is based on some important fundamental differences in organization between an avian-type nuclear pallium and mammalian sheet-like cerebral cortex. Although these differences do not preclude the evolution of comparable cognitive abilities in birds and nonhuman mammals, the idea presented here is that the differences may be important for the evolution of a human level of intelligence. Thus, while the advantages of arraying neurons into layered sheets at the surface as opposed to cell groups scattered throughout a large 3D mass might not have been evident in early members of the mammalian and reptilian lineages (were there any person present at the time to assess their relative advantages) when these patterns first emerged, my suggestion is that this arrangement in mammals possessed a greater potential for efficient expansion than did the sauropsid arrangement. In this regard, it is of interest to consider the possibility therefore that the two starting points (the reptilian and the mammalian) may have seemed trivially different at the outset (Reiner, 1993(Reiner, , 2000Reiner et al., 2005), yet the differences may have constrained the limit of what could be evolved from them.
Clearly, it is highly speculative to consider whether dinosaurs in general or any given dinosaur, for example, theropods such as troodon, could have evolved to a human intelligence level. One difficulty inherent to such an endeavor is that the conclusions reached are in the end immune to direct experimental scrutiny and thus empirical val-  (Hahn et al., 2021;Nieder, 2017;Nieder et al., 2020;, but most studies that have been done have emphasized similarities rather than differences between birds and mammals. The inefficiency should be most manifest for the projections from the layer 2/3-like sensory belt regions of DVR to the layer 5a-like neurons of dorsolateral pallium, and the projection from these layer 5a-like neurons to the layer 5b/6 neurons of arcopallium. Is any such inefficiency manifest in living birds, or do the axon distances only become a problem with expansion to a humanoid 200 cortical areas and/or 16.3 billion pallial neurons? Second and relatedly, assessing the merits of these assumptions encourages us to consider the functional consequences of the differences in "cerebral" architecture between birds and mammals. This issue may have implications for intelligent computer design, particularly if the avian design involves some unique solutions to the axon length problem. I believe thus that the merits of the notions presented here can be assessed further and doing so will provide benefit by deepening understanding of the computational power attainable with different types of forebrain organization. The approach I have taken in concluding that it might not have been possible for a dinosaur humanoid to evolve from troodon involves speculating about the axon length problem created by the expansion of the avian/dinosaur pallial design to the correspondent of a human level of 200 cortical areas, with an attendant 16.3 human-level pallial neuron abundance. As noted previously in this article, numerous studies have reported that corvids and parrots parallel primates in their intelligence, even in some regards paralleling some ape species (including chimpanzee), despite their much smaller brain size. As noted previously, the Shanahan et al. (2013) estimate of 48 separate pallial regions in pigeons leads to the conclusion that pigeons have the equivalent of about 10 cortical areas in a mammal, putting pigeons in the range of primitive living mammals such as insectivores. Since pigeon pallium has 50-70 million pallial neurons (Olkowicz et al., 2016;Ströckens et al., 2022), does this mean that a pigeon-size corvid such as the rook with about 10 times as many pallial neurons (Ströckens et al., 2022) has the equivalent of 100 or so cortical areas in its pallium? If so, does it perform the same in cognitive tasks as primates possessing 100 cortical areas and around 500 million cortical neurons, such as marmosets (Olkowicz et al., 2016). Taking this approach in evaluating the ideas presented here requires mapping of pallial areas in birds using criteria similar to those used in mammals, but it does get at the issue of the whether an avian design with a particular number of cortical areas performs less well than the mammalian design with the same number of cortical areas, and whether any such disparity grows greater with increasing numbers of "cortical" areas.
Finally, I want to again emphasize that the ideas presented here should not be construed as promoting a return to the outdated notion that mammals are inherently more intelligent than birds because they possess a laminar pallium (the neocortex) and birds are so unfortunate as to possess a nuclear pallium (the Wulst and DVR). Rather, I am suggesting that a laminar pallium may be more expansible to a human level without a significant loss of processing power than is a nuclear pallium. Clearly, however, the avian-style pallium can be expanded in size, complexity, and neuron abundance to allow the impressive, adaptive behavior and cognitive achievements seen in crows and parrots, for example (Auersperg & von Bayern, 2019;Balakhonov & Rose, 2017;Kabadayi & Osvath, 2017;Emery, 2006;Emery & Clayton, 2004a, 2004bPika et al., 2020;Sol et al., 2022). Such behaviors and cognitive achievements rival those seen in nonhuman primates, and are not seen in most mammalian species, and establish that a neocortical pallial organization is not needed for high-level behavioral sophistication (Güntürkün & Bugnyar, 2016;Guy & Staiger, 2017;Jarvis et al., 2005;Reiner, 2005Reiner, , 2009Reiner et al., 2001;Shimizu & Karten, 1991). I am, however, suggesting that the nuclear avian pallium sets a limit on the intelligence level that can be achieved, a limit that can be surpassed by a neocortical pallium-precluding a dinosaur humanoid in the sauropsid lineage, but enabling humans in the mammalian lineage.

AUTHOR CONTRIBUTIONS
This is a review article, and I am responsible for all text and figures.

CONFLICT OF INTEREST STATEMENT
The author declares no conflicts of interest.

DATA AVAILABILITY STATEMENT
This is a review article and all data used have already been published by diverse authors, and the cited articles are publicly available.

PEER REVIEW
The peer review history for this article is available at https://publons. com/publon/10.1002/cne.25458.

ENDNOTES
1 The human-like troodon descendant proposed by Russell and Seguin is often referred to as a dinosauroid, but this is a misuse of the suffixoid, which means -like. Thus, a dinosauroid is a creature that is like a dinosaur rather than like a human. For this reason, I use the term dinosaur humanoid. 2 I use the Nomenclature Forum definition of the mesopallium and hyperpallium boundaries (Reiner et al., 2004) for the reasons laid out by Montiel and Molnar (2013), Puelles (2022), Puelles et al. (2016), and Ströckens et al. (2022).