1 
 
Early origins of divergent patterns of morphological evolution on the mammal and reptile 1 
stem-lineages 2 
Running Head: EVOLUTION OF STEM MAMMALS AND REPTILES 3 
 4 
Neil Brocklehurst1*, David P. Ford2, Roger B. J. Benson2 5 
1Department of Earth Sciences, University of Cambridge, Downing Street, Cambridge, UK 6 
2Department of Earth Sciences, University of Oxford, South Parks Road, Oxford, UK 7 
* Coresponding Author. Email: nb661@cam.ac.uk 8 
 9 
 10 
 11 
 12 
 13 
 14 
 15 
 16 
 17 
 18 
 19 
 20 
 21 
 22 
 23 
 24 
ABSTRACT  25 
2 
 
The origin of amniotes 320 million years ago signalled independence from water in 26 
vertebrates and was closely followed by divergences within the mammal and reptile stem 27 
lineages (Synapsida and Reptilia). Early members of both groups had highly similar 28 
morphologies, being superficially ‘lizard-like’ forms with many plesiomorphies. However, 29 
the extent to which they might have exhibited divergent patterns of evolutionary change, with 30 
potential to explain the large biological differences between their living members, is 31 
unresolved. We use a new, comprehensive phylogenetic dataset to quantify variation in rates 32 
and constraints of morphological evolution among Carboniferous–early Permian amniotes. 33 
We find evidence for an early burst of evolutionary rates, resulting in the early origins of 34 
morphologically distinctive subgroups that mostly persisted through the Cisuralian. Rates 35 
declined substantially through time, especially in reptiles. Early reptile evolution was also 36 
more constrained compared to early synapsids, exploring a more limited character state 37 
space. Postcranial innovation in particular was important in early synapsids, potentially 38 
related to their early origins of large body size. In contrast, early reptiles predominantly 39 
varied the temporal region, suggesting disparity in skull and jaw kinematics, and 40 
foreshadowing the variability of cranial biomechanics seen in reptiles today. Our results 41 
demonstrate that synapsids and reptiles underwent an early divergence of macroevolutionary 42 
patterns. This laid the foundation for subsequent evolutionary events and may be critical in 43 
understanding the substantial differences between mammals and reptiles today. Potential 44 
explanations include an early divergence of developmental processes or of ecological factors, 45 
warranting cross-disciplinary investigation. 46 
Key Words: Amniote; Phylogeny; Rate; Constraint; Body Size 47 
 48 
 49 
INTRODUCTION 50 
3 
 
Amniotes, the terrestrialised vertebrates, are a diverse group comprising more than 51 
25,0000 living species. Their earliest fossils occur around 318 million years ago, and already 52 
include representatives of the two major subgroups that persist to the present day (Carroll 53 
1964; Reisz 1972; Mann et al. 2020): Synapsida (mammal-line amniotes) and Reptilia, or 54 
Sauropsida (the stem-lineage of reptiles, including birds; hereafter referred to as Reptilia or 55 
‘reptiles’). The earliest members of both groups were extremely similar in their general 56 
morphology, being small and superficially lizard-like insectivores with sprawling limb 57 
orientations. However, they rapidly radiated into a substantial ecomorphological diversity, 58 
including diversification of diets (Sues & Reisz 1998, Brocklehurst & Benson 2021), body 59 
sizes (Laurin 2004; Reisz & Fröbisch 2014; Brocklehurst 2016; Brocklehurst & Brink 2017; 60 
Brocklehurst & Fröbisch 2018, Brocklehurst et al. 2020), habitat use (e.g. arboreality; 61 
Spindler et al. 2018; Mann et al 2021), and diel activity patterns (Angielczyk & Schmitz 62 
2014, Ford & Benson 2019). Their success has been attributed to a number of evolutionary 63 
innovations, including musculoskeletal adaptations that freed the skull from its role in lung 64 
ventilation, allowing greater skull versatility (Frazetta 1968; Janis & Keller 2001), the 65 
evolution of temporal fenestration facilitating muscle attachment (Frazetta 1968, Wernberg 66 
2019; Abel & Wernberg 2021), and the evolution of the amniotic egg (Romer 1957; Carroll 67 
1970). 68 
Early amniotes provide a classic example of diversification following adaptive zone 69 
invasion, and various studies have sought to characterise macroevolutionary patterns during 70 
this transition, with suggestions of little substantial change in rate or mode of morphological 71 
evolution of body size or general anatomy at the origin of amniotes (Laurin 2004; Ruta et al. 72 
2006, 2018), but substantial increases in functional disparity of the feeding apparatus 73 
(Anderson & Friedman 2013) and in rates of tooth and jaw evolution (Brocklehurst & Benson 74 
2021). However, understanding of the early radiation within amniotes is less well-75 
4 
 
characterised. Analyses so far have been conducted at various phylogenetic scales, including 76 
in larger analyses of tetrapod evolution that contain a more limited sampling of early 77 
amniotes (e.g. Laurin 2004; Ruta et al. 2006, 2018; Anderson & Friedman 2013), as well as 78 
restricted  examinations of early amniote subgroups (Brocklehurst 2016, 2017; Brocklehurst 79 
& Brink 2017; Romano et al. 2017, 2018; MacDougal et al. 2019). Studies have also focused 80 
on different portions of the anatomy, including limbs (Ruta et al. 2018), jaws and teeth 81 
(Anderson & Friedman 2013; Brocklehurst & Benson 2021), vertebrae (Jones et al. 2018, 82 
2020), and body size (Laurin 2004). However, thus far there has been no study of 83 
macroevolutionary patterns during the origin and early radiation of amniotes including a 84 
broad selection of all clades, allowing direct comparison of the evolutionary patterns within 85 
the major lineages, and across many anatomical regions. 86 
 87 
Discrete character state matrices provide observations of morphological variation 88 
from across the skeleton that may be used for large-scale macroevolutionary analyses. We 89 
present a new phylogenetic dataset, substantially expanded from the most recent phylogenetic 90 
assessment of early amniote evolution (Ford and Benson 2020), including species and 91 
relevant anatomical variation from across all clades spanning the Carboniferous until the end 92 
of the early Permian. We use this to assess rates of evolution and evolutionary constraints 93 
during the earliest radiation of amniotes across their anatomy and within different partitions, 94 
examining differences between early synapsids and early reptiles. 95 
 96 
MATERIALS AND METHODS 97 
Dataset 98 
We analyse a new phylogenetic dataset of early amniotes, focused on coverage of late 99 
Carboniferous (Pennsylvanian) and Early Permian (Cisuralian) taxa, and lineages that 100 
5 
 
survived into the middle Permian (Guadalupian), including Therapsida and Neodiapsida. We 101 
attempted comprehensive coverage of phylogenetically-informative characters from previous 102 
studies and our own observations (e.g. Benson 2012, Modesto et al 2014), expanded from the 103 
analysis of Ford & Benson (2020) to achieve a broader sample of Paleozoic amniote lineages. 104 
To this end, we added 31 new taxa, mostly pelycosaurian-grade synapsids, moradisaurine 105 
captorhinids and acleistorhinid parareptiles. 72 additional characters were added, mostly 106 
sourced from Benson (2012) and Modesto et al (2014). The final dataset contains 98 taxa and 107 
366 characters (Supplementary Data 1 and 2). 108 
Our study also includes analyses of a dataset of 144 dental traits in 534 taxa, taken 109 
from Brocklehurst & Benson (2021). This was included to evaluate variation in dental traits 110 
because it contains a greater sampling of dental characters from both jaws and the palate, 111 
being designed to investigate macroevolutionary patterns within feeding apparatus. It also 112 
contains a greater sampling of taxa, both within the interval of study and subsequent times 113 
until the Early Triassic, allowing analysis over a longer time duration than available for our 114 
primary matrix (Supplementary Data 3). 115 
 116 
Fossilised Birth Death Analysis 117 
We used Bayesian phylogenetic inference under a relaxed MkV model of character 118 
state evolution with a Fossilised Birth Death (FBD) tree prior (Heath et al 2014) to infer a 119 
time-scaled phylogeny and rates of character state evolution. To account for the uncertainty 120 
in the time of the first appearances, the ages of taxa were represented by a uniform 121 
probability distribution covering the full uncertainty of the age of the formation or 122 
assemblage zone in which they first appear (See Supplementary Data 20 for origin of 123 
formation ages). Net speciation rate (diversification) was drawn from a uniform prior, with 124 
net extinction rate (turnover) and relative fossilisation rate drawn from beta priors. Temporal 125 
6 
 
variation in these parameters was not modelled independent gamma rates model was 126 
employed to account for rate heterogeneity between branches (an uncorrelated clock model 127 
where rates are drawn from a gamma distribution). Rate heterogeneity between characters 128 
was also modelled as a gamma distribution. The analysis was carried out with two runs 129 
containing four chains for 50 million generations, sampling every 1000, with 25% of trees 130 
discarded as burn-in. The maximum clade credibility tree was used as the phylogenetic 131 
framework for subsequent analyses. The analysis was implemented in MrBayes 3.2.6 132 
(Ronquist & Huelsenbeck. 2003). 133 
 134 
Analysis of Rates 135 
Rates of character change along each branch were drawn from the results of the FBD 136 
analysis (Supplementary Data 4). Variation in the rates of evolution through time was 137 
assessed by time slicing the tree at intervals of one million years between 320 and 272 mya 138 
(from the origin of amniotes until the end of the Cisuralian). The rates of all branches 139 
crossing each time slice (not including non-amniote outgroups) were collated, and the median 140 
rate of each time slice was calculated. In order to assess long-term trends in rate variation, a 141 
Loess regression curve was fitted to the median rate values through time.  142 
 143 
Analysis of Constraint 144 
Variation in the strength of evolutionary constraint among lineages was assessed by 145 
comparing patristic distances and morphological dissimilarities between pairs of taxa, 146 
expanding on a procedure designed by Brocklehurst et al. (2021) to assess character state 147 
saturation, or exhaustion: the point where further evolutionary change in morphology (i.e. 148 
increasing patristic morphological distance) no longer results in an increase in the differences 149 
between taxa (i.e. morphological dissimilarity, or disparity), but instead explores a pre-150 
7 
 
established character state space, with a high prevalence of homoplasy. We indexed the 151 
morphological dissimilarity between pairs of taxa as the proportion of character scores that 152 
differ between them, calculated in the R v3.6.1 (R core team 2019) in the package Claddis 153 
(Lloyd 2016) using the MORD distance metric. Evolutionary change (patristic morphological 154 
distance) is represented by the total phylogenetic branch length between a pair of taxa, 155 
representing the number of character state changes that evolved since divergence from their 156 
common ancestor. To calculate this, the character/taxon matrix was reanalysed in MrBayes 157 
using an a Mkv model of character state evolution, with no information on taxon ages, 158 
constraining the topology to that found by the FBD analysis (for our primary matrix) or a 159 
composite tree representing consensus from the literature (for our additional dental matrix; 160 
see Brocklehurst & Benson [2021]). This Mkv analysis produced a phylogeny in which 161 
branch lengths correspond to the inferred amount of morphological character state change 162 
(Supplementary Data 5). The summed branch lengths between pairs of taxa were then used as 163 
patristic morphological distances, extracted using the R package adephylo (Jombart et al 164 
2010).  165 
In general, morphological dissimilarity should increase with evolutionary state 166 
changes (i.e. with increasing patristic morphological distance). However, this increase begins 167 
to asymptote at higher patristic distances (Wagner 2000), indicating the a lack of further 168 
exploration of novel  character state space. This occurs because homoplastic state changes 169 
and reversals can cause increases in similarity, and homoplasy becomes more frequent with 170 
increasing patristic distance under constrained evolution (Brochu 1997, Wagner 2000). This 171 
results in character state saturation (Foote 1994), or exhaustion (Wagner 2000), whereby 172 
further increases in patristic morphological distance between taxon pairs does not, on 173 
average, lead to greater morphological dissimilarity between them. Character state saturation, 174 
indicated by the asymptote of the relationship between morphological dissimilarity and 175 
8 
 
patristic morphological distance, occurs at lower morphological dissimilarity when 176 
constraints are strong, and higher dissimilarity when constraints are weak (Wagner 2000).  177 
Individual groups were assessed for significant increases or decreases in constraint 178 
using the procedure of Brocklehurst et al. (2021), in which a Michaelis-Menten curve was fit 179 
to the comparisons of the patristic morphological distances and pairwise morphological 180 
dissimilarities between all pairs of taxa within that clade, the Vmax (asymptote) parameter of 181 
that curve being used to represent the point of character state saturation. The significance of 182 
differences in Vmax between portions of the phylogeny was evaluated by comparison to 183 
expectations given a uniform model of evolution, as described in Brocklehurst et al. (2021). 184 
This was implemented by simulating null character/taxon matrices under an equal rates 185 
model, with missing data scores added in the same location as in the empirical dataset. Null 186 
morphological dissimilarities between the taxa were calculated from these matrices as 187 
described above, which were compared to the patristic distances again by fitting a Michaelis-188 
Menten curve, showing whether the clade under study reached character state saturation at a 189 
higher or lower level than in the null simulations. Character state saturation was assessed in 190 
both reptiles and synapsids, restricting the comparisons using the primary matrix to pairs of 191 
taxa that diverged within the interval of time under study: the Carboniferous–early Permian. 192 
We omitted younger branches because they were incompletely sampled and had only been 193 
included to ensure coverage of branches that originated in the early Permian, including some 194 
that continued into later intervals. However, our analysis of the dental dataset of Brocklehurst 195 
& Benson (2021) (Supplementary Data 6) extends up to the early Triassic allowing insights 196 
into important later events such as the diversification of Therapsida, ankyromorph 197 
parareptiles and Neodiapsida, at least for dental characters. 198 
We expand on this method here to allow detection of variation in the strength of 199 
constraint in the absence of prior hypotheses regarding the phylogenetic location of shifts. 200 
9 
 
This was done by applying the method to every node in the phylogeny, comparing the 201 
patristic morphological distances and pairwise dissimilarities as described above, by fitting a 202 
Michaelis Menten curve to estimate the Vmax parameter and 84% confidence intervals around 203 
it. We then compared this to the null distribution resulting from analysis of null 204 
character/taxon matrices resulting from 1000 iterations of our simulation approach. 205 
Significance was determined when the 84% confidence interval of a node’s observed Vmax 206 
value lay entirely above the 84% quantile of the null simulations (significant release in 207 
constraint), or entirely below the 84% quantile of the null simulations (significant 208 
strengthening of constraints). We used 84% intervals because they are expected to overlap 209 
95% of the time when two distributions are statistically identical, therefore representing a 210 
significance threshold of 0.05 (Payton et al 2003). In contrast, two statistically identical 95% 211 
intervals will overlap 99% of the time, resulting in increased frequency of false negatives. 212 
The entire process is carried out in R using custom code (Supplementary Data 7 and 8) 213 
written using functions from the packages Adephylo (Jombart et al 2010), Claddis (Lloyd 214 
2016), and Phytools (Revell 2012). 215 
 216 
Analyses of Character Partitions 217 
Constraints within anatomical partitions were investigated by analysing subsets of the 218 
full character list of each analysis, representing (i) the skull (including lower jaw and 219 
mandible), (ii) postcranium (iii) snout (antorbital region of the skull, not including palate) (iv) 220 
temporal (postorbital region of the skull) and (v) dentition (including palatal dentition; and 221 
supplemented by analysis of the more extensive dataset of dental traits from Brocklehurst & 222 
Benson 2021) (Supplementary Data 9-13). For analysing rates within the partitions, 223 
characters from each to were subjected to an FBD analysis, with node ages and topology 224 
constrained to those identified by the FBD analysis of the whole dataset. This produces a tree 225 
10 
 
identical to the MCC tree from analysis of the whole dataset, but where the rate values 226 
represent only those of the character partition (Supplementary data 14-18). For analysis of 227 
constraints, characters from each partition to were subjected to an undated Bayesian analysis 228 
constraining the tree to that identified by the FBD analysis of the whole dataset. This 229 
produces a tree whose branch lengths represent only character changes within the relevant 230 
partition. Rates of evolution and character state saturation were then assessed as described 231 
above. 232 
 233 
Analysis of Body Size 234 
Differences in patterns of trait evolution between reptiles and synapsids may result 235 
from the macroevolutionary effects of large body sizes, which evolved frequently among 236 
early synapsids but rarely among early reptiles (Modesto et al. 2015; Brocklehurst & Brink 237 
2017; Brocklehurst & Fröbisch 2018). To evaluate this, taxa were assigned to one of four size 238 
categories, each representing an order of magnitude: (small:<1kg, medium:1-10kg, large: 10-239 
100kg, very large: 100-1000kg) (Supplementary Data 19). Using discrete categories of body 240 
size is less precise than estimates of size using continuous measurements (e.g. Alroy 1998; 241 
Campione & Evans 2012; Brocklehurst & Brink 2017; Benson et al. 2018). Nevertheless, we 242 
use discrete body size categories here because they allow the inclusion of the maximum 243 
number of taxa, including those too fragmentary to make precise mass estimates. Constraint 244 
within each size class was assessed as described above, comparing pairs of taxa within each 245 
size category, limiting comparisons to pairs of taxa that diverged during the Carboniferous–246 
early Permian. 247 
In order to examine how size evolution varied between synapsids and reptiles, four 248 
models of discrete character evolution were fit to the phylogeny and body size categories: 249 
Single regime, equal rates (all possible transitions between character state have a single rate, 250 
11 
 
and the rate is consistent between reptiles and synapsids); Single regime, all rates different 251 
(all possible transitions between character state may a have a different rate, but the rates are 252 
consistent between reptiles and synapsids); Multi regime, equal rates (all possible transitions 253 
between character state have a single rate, but the rate varies between reptiles and synapsids); 254 
Multi regime, all rates different (all possible transitions between character state may a have a 255 
different rate, and the rates are vary between reptiles and synapsids). The model fitting was 256 
carried out using the fitMultiMk function in phytools (Revell 2015). The fit of the models to 257 
the data was assessed using the Akaike weights.  258 
 259 
RESULTS 260 
 261 
Phylogeny 262 
Our phylogenetic analysis returns a tree topology that is consistent with that found by 263 
Ford & Benson (2021), in spite of the addition of characters and taxa (Fig. 1; Fig S2). 264 
Parareptiles are found as the sister to neodiapsids and varanopids are found as reptiles, 265 
suggesting that support for this contentious phylogenetic hypothesis (e.g. Benoit et al. 2021; 266 
Bazzana et al. 2021) remains high even given a larger sample of early synapsids. The 267 
addition of further pelycosaurian-grade synapsids produced results broadly consistent with 268 
consistent with recent analyses of this grouping (Brocklehurst & Fröbisch 2018, Maddin et al. 269 
2020; Berman et al. 2020). The only noticeable discrepancy is that Eocasea, Callibrachion 270 
and Datheosaurus are found as outgroups to other caseasaurs (Eothyrididae and Caseidae) 271 
rather than being within caseids (as found by Resiz & Fröbisch 2014, Brocklehurst et al. 272 
2016, Berman et al. 2020). 273 
 274 
Analysis of Ratesz 275 
12 
 
Rates were highest during the earliest history of amniotes, decreasing through the late 276 
Carboniferous until the middle of the early Permian (Cisuralian) (Fig.1, 2a). Both synapsids 277 
and reptiles exhibit high evolutionary rates that decreased through time during the late 278 
Pennsylvanian/earliest Cisuralian. However, rates of evolution in synapsids remain about 279 
twice as high as those of reptiles through the latter half of the Cisuralian (Fig. 2b). Among 280 
subclades, rates are highest in Eupelycosauria (among synapsids), and during the early 281 
divergences of varanopids and parareptiles (among reptiles) (Fig. 1). 282 
 283 
Analysis of Constraints 284 
A relaxation of constraint relative to null expectations is observed at the base of 285 
Synapsida, followed by their strengthening within individual lineages: caseids, 286 
eupelycosaurs, and sphenacodontians (within Eupelycosauria) (Fig. 3a). Constraints 287 
strengthen at the base of Reptilia but becomes relaxed around the earliest divergences of 288 
Diapsida. Subsequent strengthening of constraints is observed within the mycterosaurine 289 
varanopids and parareptiles. The overall pattern of relaxed constraints and elevated rates 290 
during early amniote evolution, followed by slowdowns and increased constraints, are 291 
consistent with models proposed by Simpson (1953) for phenotypic evolution during 292 
adaptive radiations: rapid evolution between peaks in the adaptive landscape and the lineages 293 
diverge into different regions of ecospace, followed by subsequent reductions in rate as 294 
niches are filled and constraint increases within the adaptive optima.  295 
Constraints played a more important role during early reptile evolution than in early 296 
synapsids. While both groups experienced initially relaxed constraints early in their 297 
evolution, reptiles reached character state saturation before the end of the Cisuralian; 298 
evidenced by asymptoting of the relationship between patristic distance and morphological 299 
dissimilarity (Fig 3b, Table 1). Synapsids, on the other hand, did not reach character state 300 
13 
 
saturation, and were therefore continuing to explore new areas of morphospace (Fig. 3b). 301 
While the interquartile range and media of the morphological dissimilarities between taxa are 302 
similar for reptiles and synapsids, the maximum dissimilarities observed are considerably 303 
higher in synapsids, indicating their access to a larger area of morphospace (Fig 3c). 304 
 305 
Variation between anatomical partitions. 306 
 307 
 Reptiles and synapsids differ in how patterns of evolution are expressed ampngdifferent 308 
anatomical partitions. Both lineages xhibit relaxed constraints on skull evolution, in particular 309 
for the snout (Fig 4a,b). In contrast, constraints in the temporal region of the skull, which 310 
houses the jaw closing muscles, and signifies important functional variation, are significantly 311 
relaxed in reptiles but not synapsids (Fig 4c). This gave rise to higher median and maximum 312 
values of morphological dissimilarity in this region among reptiles (Fig S3c). Synapsids 313 
experienced elevated rates of skull evolution during their earliest evolution, which declined 314 
through the late Carboniferous before recovering slightly during the late Cisuralian (Fig 5a). 315 
In contrast, rates of skull evolution in reptiles remained low throughout the study interval.  316 
Reptiles show high early rates of postcranial evolution compared to those of synapsids 317 
(Fig 5b). Nevertheless, postcranial rates declined through time and exhibit significantly high 318 
constraints in reptiles compared to synapsids, suggesting that high early rates did not result in 319 
proliferation of a wide range of postcranial morphologies in reptiles (Fig 4d). Synapsids, 320 
experienced a significant relaxation of constraints on postcranial evolution (Fig 4d), 321 
suggesting that, although they evolved more slowly, they potentially acquired a wider 322 
disparity of postcranial morphologies (Fig S3d).  323 
Dental traits show an early release of constraints in both reptiles and synapsids. 324 
However, while reptiles retained these relaxed constraints within Diapsida and Neoreptilia, 325 
14 
 
synapsids experienced a strengthening of constraints within eupelycosaurs (Fig S3a). This 326 
was confirmed in our analyses of the larger dental dataset of Brocklehurst & Benson (2021), 327 
which could be analysed over a longer study interval and provides evidence for a subsequent 328 
relaxation of constraints on dental evolution in the synapsid subgroup Therapsida (Fig S3b).  329 
 330 
 331 
Constraint at different body sizes 332 
Small and medium sized amniotes were found to evolve under greater constraint than large 333 
and very large taxa, reaching character state saturation at a lower morphological dissimilarity 334 
(Fig. 6b). Interestingly, taxon-pairs within all three size classes are found to evolve under 335 
greater constraint than expected from null simulations (Table 1). This suggests that 336 
morphological innovation among early amniotes was predominantly associated with 337 
evolutionary changes between size categories. Overall, therefore, taxa within different size 338 
classes occupy different regions of morphospace, with larger taxa occupying larger regions of 339 
morphospace (Fig S6). 340 
Although reptiles and synapsids show different patterns of body size evolution (see 341 
below), they nevertheless show similar patterns of overall morphological constraint within 342 
the size categories. Large and very large taxa exhibit relaxed constraints in both synapsids 343 
and reptiles, although early synapsids include more taxa of these sizes than do reptiles (Fig 344 
7c, Table 1).  345 
We also find evidence for different patterns of body size evolution among synapsids 346 
compared to those in reptiles (see also Modesto et al. [2015]; Brocklehurst 2021). Evolution 347 
of the discrete character representing body size best fits a model where the two lineages are 348 
subject to different evolutionary regimes (Table 2). Within synapsids, only transitions from 349 
smaller size categories to larger have positive rates; rates of transition from larger size 350 
15 
 
categories to smaller were all zero (Fig. 7). This indicates an evolutionary  trend towards 351 
larger body sizes in synapsids. Reptiles, on the other hands, have positive rate values for 352 
transitions in both directions, and in some cases the rate of transition from larger to smaller 353 
size categories is higher than from smaller to larger. 354 
 355 
DISCUSSION 356 
The evolution of full terrestriality, at the origin of amniotes, was a major event in 357 
vertebrate evolution, providing insights into macroevolutionary patterns during ecospace 358 
invasion. Moreover, the evolutionary divergences among early amniotes gave rise to the 359 
mammal-line (Synapsida) and reptile-line (Reptilia), which persist to the present day and 360 
show stark differences in their morphology, ecology, and biology. Previous studies indicated 361 
variation in rates and constraint coinciding either with the origin of amniotes or their 362 
subsequent diversification into different areas of terrestrial ecospace. This has been shown, 363 
for example, during body size evolution (Laurin 2004; Reisz & Fröbisch 2014; Brocklehurst 364 
2016; Brocklehurst & Brink 2017; Brocklehurst & Fröbisch 2018, Brocklehurst et al. 2020), 365 
and the evolution of jaw (Anderson et al. 2013), tooth (Brocklehurst & Benson 2021) and 366 
limb morphology (Ruta et al 2018). However, macroevolutionary patterns during this 367 
transition have been unclear due to the variation in taxonomic scope of these studies and 368 
regions of anatomy analysed. Our study represents the first analysis that samples broadly 369 
across early members of the amniote crown-group and including morphological variation 370 
from across the whole skeleton. 371 
We find evidence for an early episode of high evolutionary rates across the skeleton, 372 
coupled with relaxed constraints on cranial (especially snout and dental) evolution. This is 373 
consistent with the early origins of a wide set of morphologically and ecologically distinctive 374 
amniote subclades by the late Carboniferous, including herbivorous edaphosaurids (Sues & 375 
16 
 
Reisz 2000), macropredatory sphenacodontians (Fröbisch et al. 2010; Brocklehurst & Brink 376 
2017), arboreal protorothyridids (Mann et al. 2021), and many other groups. These findings 377 
contradict those of Ruta et al. (2006), who suggested that amniote origins did not give rise to 378 
an increase in the rate of character state evolution. This may be an artefact resulting from 379 
under-sampling of amniote taxa and characters in the dataset of Ruta et al. (2006), who 380 
included only a very small sample of crown amniotes, from the reptile line only. 381 
Nevertheless, future studies should examine whether the elevated rates and subsequent 382 
decline observed here in early amniotes merely represents a subset of a longer-term decline in 383 
rates across tetrapods.  384 
Our findings of relaxed constraints on cranial, snout and dental evolution in the 385 
earliest amniotes are consistent with the hypothesis that diet-related cranial variation was an 386 
important axis of phenotypic diversification during their initial radiation (e.g. Janis & Keller 387 
2001; Anderson & Friedman 2013; Brocklehurst & Benson 2021). We also show that rates of 388 
evolution were also elevated for other anatomical regions, not strictly limited to dietary or 389 
craniodental diversification. Nevertheless, those regions (postcrania, and the temporal region 390 
of the skull) exhibit different patterns of variation in constraint between reptiles and 391 
synapsids (Fig. 4), highlighting the divergent paths to morphological diversification that were 392 
taken by these groups. 393 
The initial diversification of reptiles appears to have been focused on the temporal 394 
region. This is consistent with the qualitative observation that early reptiles exhibit 395 
considerable evolutionary versatility of temporal fenestration, contrasting with more 396 
conserved temporal anatomy in synapsids (Piñeiro et al. 20212; MacDougal & Reisz 2014; 397 
Haridy et al. 2016; Ford & Benson 2020). Diversification of temporal fenestration among 398 
early reptile groups likely corresponds to variation in muscle attachment and jaw function 399 
(Frazzetta 1968; Wernberg 2019, Abel & Wernberg 2021), and may therefore reflect an early 400 
17 
 
diversification of cranial function. Reptiles also maintain relaxed constraints on their 401 
dentition, whereas eupelycosaurian synapsids experience a strengthening of the constraints. 402 
This may result from differences in the evolvability of palatal dentition between synapsids 403 
and early reptiles. Reptiles exhibit a great diversity in the arrangement, size, density, and 404 
patterns of loss of the palatal dentition, both today and in the past (Matsumoto and Evans 405 
2017). Such variation is also present, to some extent, in caseasaurian synapsids (Brocklehurst 406 
et al. 2016). However, eupelycosaurian synapsids show much less variability, with a trend to 407 
simplification and ultimately loss of palatal teeth long before the origin of mammals 408 
(Matsumoto and Evans 2017). This may have been compensated by a relaxation of 409 
constraints on the marginal dentition along the line leading to mammals, which is evident 410 
among the middle Permian divergences of therapsids. (Fig S3b). Increases in the evolutionary 411 
versatility of marginal dentition among synapsids culminated in the development of strongly 412 
heterodont and functionally differentiated marginal dentitions as a central innovation of later 413 
cynodonts, including mammals (Compton & Jenkins 1968; Luo et al. 2015).  414 
Postcranial data indicate a decoupling of rates and constraints. Reptiles show high 415 
early rates coupled with significantly increased constraint suggesting that they rapidly 416 
explored a relatively small postcranial character state space. In contrast, early synapsid 417 
postcrania evolved at lower rates for much of the Carboniferous and early Permian, but under 418 
significantly relaxed constraints that allowed them to gradually explore a much larger 419 
character state space. Our findings are therefore potentially consistent with the limited 420 
previous studies of synapsid postcranial variation, which found evidence for high disparity of 421 
humerus shape and heterogeneity between vertebral regions in therapsids (Jones et al. 2018; 422 
Lungmus & Angielczyk 2019).  423 
Synapsid evolution is further distinct from that of reptiles in that synapsids rapidly 424 
attained large body sizes during their early history, whereas reptiles did not. Ancestral 425 
18 
 
character state mapping suggests a small-bodied ancestor of the amniote crown-group (< 1 426 
kg) (Fig 7a), from which multiple synapsid lineages independently evolved large body sizes 427 
exceeding 40 kg before the end of the Carboniferous (Reisz & Fröbisch, 2014; Brocklehurst 428 
& Brink, 2017; Brocklehurst & Fröbisch, 2018). In contrast, reptiles did not reach such sizes 429 
until the latest Cisuralian origin of moradisaurine captorhinids (Brocklehurst, 2016). The 430 
analysis of body size evolution presented here should be treated with caution due to the low 431 
resolution of the size data, but it does demonstrate that synapsid body size was evolving 432 
under a distinct regime to that of reptiles, and potentially exhibited a trend of increasing body 433 
size that was absent in reptiles. 434 
 It has been suggested that the apparent increase in body size may have driven the 435 
apparent greater diversity of synapsids during the Paleozoic, due to biases in either 436 
preservation or collection: larger synapsids may be easier to find or have greater fossilisation 437 
potential than smaller, more fragmentary reptiles (Modesto et al. 2015; Brocklehurst 2021). 438 
The fossil reptiles named from the Carboniferous are smaller, but more complete than the 439 
synapsids (Modesto et al. 2015). This potentially indicates that only the most complete 440 
reptiles are collected or described, whereas synapsid material may be considered informative 441 
even when it is more fragmentary, due to the larger body size of synapsids (Modesto et al. 442 
2015). However, it is also possible that this signal is genuine, and reflects the earlier adoption 443 
of herbivory and carnivory in synapsids (Modesto et al. 2015), reflecting a wider pattern of 444 
divergence along ecological lines in early amniotes.  445 
Irrespective of patterns of species diversification, larger body size appears to be 446 
related to reduced constraint in morphological evolution, both in reptiles and synapsids (Fig. 447 
6b,c). Therefore, the fact that synapsids show an early trend towards larger body size, 448 
reaching larger sizes earlier and more frequently than the reptiles, may have permitted the 449 
greater relaxation of constraints observed in early synapsids. A greater range of body sizes, as 450 
19 
 
observed in synapsids, has been linked to greater functional diversity (Woodward et al. 2005; 451 
Rooney & McCann 2012). Larger body sizes allow access to a different range of ecotypes, 452 
including macro-predation and high fibre herbivory (Clauss & Hummel 2005; Müller et al. 453 
2013; Brocklehurst & Brink 2017), permitting further diversification within these distinct 454 
regions of ecomorphospace. Moreover, the fact that synapsids rarely underwent reversals to 455 
small body sizes could result in fewer homoplasies among size-dependent characters, and 456 
therefore weaker apparent constraints on morphological evolution. 457 
Early attainment of large body size is particularly relevant to the release in constraints 458 
on postcranial evolution observed in synapsids, but not in reptiles. Scaling relationships 459 
imply that the stresses experienced by the skeleton are relatively higher in larger-bodied 460 
species (Stanley 1973; Biewener 1982). Thereby, ecological specialisation may require more 461 
substantial postcranial specialisation among large-bodied species to exploit new niches. 462 
Studies in mammals support these observations, demonstrating greater diversity within 463 
different locomotor modes at larger sizes (Weaver & Grossnickle 2020), and greater 464 
distinction between locomotor types in larger taxa (Jenkins 1974; Jenkins & Parrington 1976; 465 
Runestead & Ruff 1995). The relaxed evolutionary constraints in the synapsid postcranium 466 
may provide an explanation of why large body size evolved multiple times independently 467 
among early synapsids, but not among early reptiles; synapsids’ access to a wider region of 468 
postcranial morphospace allowed the greater postcranial specialisation necessary for 469 
ecological specialisation at large sizes. Alternatively, large body size may have evolved in 470 
synapsids for other reasons (e.g. ecological), and necessitated postcranial specialisations that 471 
are detected here as a release in constraint on postcranial evolution.  472 
 Early events in amniote evolution, documented here, set the stage for the origins of 473 
major groups that comprise most of the extant diversity of land vertebrates. In particular, 474 
Neodiapsida (including the reptile crown-group; ‘Neo’ in Fig. 1) and Therapsida (including 475 
20 
 
the mammalian crown-group; ‘The’ in Fig. 1) comprise the bulk of amniote diversity after the 476 
early Permian and have highly distinct anatomy compared to their predecessors (Rubidge & 477 
Sidor 2001). These groups evolved substantial morphological disparity (Ruta et al. 2013a,b; 478 
Ezcurra  & Butler 2018; Grunert et al. 2019; Lungmus & Angielczyk 2019). However, the 479 
branches leading to them, which span most of the our early Permian study interval, show no 480 
evidence of high rates of evolution. In fact both lineages represent evolutionary slowdowns 481 
relative to the ‘backbones’ of the reptile and synapsid phylogenies, demonstrating that high 482 
evolutionary rates are not required to explain the origins of these groups. This raises the 483 
possibility that distinctive traits of both neodiapsids and therapsids assembled gradually 484 
throughout the early Permian (Cisuralian) during a cryptic and poorly-sampled interval of 485 
evolutionary history, before their rise to high abundances during latter intervals of Permian 486 
and the Triassic. However, both groups exhibit a long interval in which direct evidence for 487 
rates of accumulation of their derived characters is entirely missing the fossil record (late 488 
Pennsylvanian–latest Cisuralian). Therefore, future fossil discoveries are required to test 489 
hypotheses of their evolutionary patterns and could demonstrate, in reality, their traits 490 
appeared abruptly, either late or early in this unsampled time window. This is particularly 491 
relevant to discussions on the origin of therapsids, where it has been suggested that the 492 
interval known as Olson’s Gap (late Kungurian-Roadian, latest Cisuralian-earliest 493 
Guadalupian) (Lucas and Heckert 2001; Liu et al. 2009; Brocklehurst 2020) was an important 494 
evolutionary interval and that better sampling at this time would shed light on would shed 495 
light on the origins of therapsid anatomy (Abdala et al. 2008; Liu et al. 2009). Our results 496 
show that current data are also consistent with a protracted origin, and that a great expansion 497 
of fossil evidence throughout the early Permian could be required to fully test this.   498 
Variation between macroevolutionary patterns among different anatomical partitions 499 
for early synapsids and early reptiles may have underpinned the different evolutionary 500 
21 
 
trajectories of reptile- and mammal-line amniotes, and may ultimately have resulted in the 501 
clear disparities between mammals and reptiles today. Our results potentially imply that a 502 
deep divergence in patterns of evolutionary modularity (Vermeej 1973; 1973; Wagner and 503 
Altenberg 1996), due to either developmental or ecofunctional drivers, might potentially 504 
explain the different paths that reptiles and synapsids have taken during their ecological 505 
diversification. For example, we provide evidence for relaxation of constraints on temporal 506 
evolution i among early reptiles (Fig. 4). This is congruent with variation seen among extant 507 
reptiles, which exhibit substantial disparity of temporal morphology and of regions associated 508 
with jaws muscular and articulation (Watanabe et al. 2019; Rhoda et al. 2020), as well as 509 
variation in the degree of cranial kinesis and the location of articulation points. This ranges 510 
from relatively akinetic skulls in crocodiles, turtles and rhynchocephalians (Preuschoft & 511 
Witzel 2002; Ferreira et al. 2020) to mobility of the frontoparietal suture, quadrate and palate 512 
in many squamates, extreme forms of kinesis in snakes (Arnold 1989; Metzger 2002; Rhoda 513 
et al. 2020), and kinesis of the beak relative to the braincase in birds (Bout & Zwiers 2001). 514 
Variability in kinesis and articulation is noticeably more limited, or absent, in mammals.  515 
Relaxation of postcranial evolution is shown here among early synapsids (Fig. 4). 516 
This is congruent with considerable variation in the postcranial morphology of modern 517 
mammals and their therapsid ancestors, including greater variability and distinction between 518 
modules in the vertebral column in mammals compared to reptiles (Arnold et al. 2017; Jones 519 
et al. 2018; Arnold 2021), and considerable increases in synapsid humeral disparity from the 520 
middle Permian onwards (Lungmus & Angielczyk 2019). Our findings therefore suggest that 521 
an initial release in postcranial evolution occurred at the origin of synapsids, with subsequent 522 
increases occurring in therapsids and among mammals (Jones et al 2018; Lungmus & 523 
Angielczyk 2019). 524 
 525 
22 
 
CONCLUSIONS 526 
Synapsids and reptiles in the present day represent two highly divergent lineages, with 527 
substantial differences in their morphology and physiology. During their earliest divergences 528 
in the Carboniferous, however, they were morphologically and ecologically very similar, both 529 
represented by small insectivores with a superficially ‘lizard-like’ body form. These 530 
superficial similarities mask a deep evolutionary divergence in rates and modes of between 531 
synapsids and reptiles, documented by our analyses. Analysis of constraint without a priori 532 
assumptions on where regime shifts occur identify fundamental differences in patterns of 533 
evolution between the two lineages; reduced constraint in the temporal region of the skull on 534 
the reptile line, and reduced constraints in postcranial evolution on the synapsid line, which 535 
may potentially be linked to a trend on increasing body size. This separation of evolutionary 536 
patterns during the earliest divergence between synapsids and reptiles may be fundamental to 537 
understanding the differences between these groups throughout their history, including stark 538 
differences between members of these groups that are present today. Our observations of a 539 
deep divergence of macroevolutionary modalities raises the possibility of a deep divergence 540 
of either developmental processes or ecological factors very early on the mammal and reptile 541 
lines. Further studies of morphological evolution, spanning the subsequent intervals of 542 
amniote evolution are required to confirm this possibility. Moreover, developmental and 543 
ecological or functional studies are also required to test the mechanisms that may have given 544 
rise to this macroevolutionary divergence. Amniote origins and the long-term differentiation 545 
of mammal and reptile phenotypes therefore provide a promising avenue for cross-546 
disciplinary investigation in evolutionary research.   547 
 548 
ACKNOWLEDGEMENTS 549 
23 
 
We would like to thank Graeme Lloyd for assistance with Claddis, and Yara Haridy for 550 
helpful discussion. Reviews by Kenneth Angielczyk, Peter Wagner and an anonymous 551 
reviewer greatly improved this paper. This research was funded by the European Union’s 552 
Horizon 2020 research and innovation program 2014–2018 under grant agreement 677774 553 
(European Research Council [ERC] Starting Grant: TEMPO) awarded to RBJB. 554 
 555 
REFERENCES 556 
Abdala, F., Rubidge, B. S. & Van Den Heever, J. (2008) The oldest therocephalians 557 
(Therapsida, Eutheriodontia) and the early diversification of Therapsida. 558 
Palaeontology, 51, 1011-1024 559 
Abel, P., & Werneburg, I. (2021). Morphology of the temporal skull region in tetrapods: 560 
research history, functional explanations, and a new comprehensive classification 561 
scheme. Biological Reviews, In Press. 562 
Alroy, J. (1998). Cope’s rule and the dynamics of body mass evolution in North American 563 
fossil mammals. Science, 280, 731-734 564 
Anderson, P. S., Friedman, M., & Ruta, M. (2013). Late to the table: diversification of 565 
tetrapod mandibular biomechanics lagged behind the evolution of terrestriality. 566 
Integrative & Comparative Biology, 53, 197-208 567 
Angielczyk, K. D., & Schmitz, L. (2014). Nocturnality in synapsids predates the origin of 568 
mammals by over 100 million years. Proceedings of the Royal Society B, 281, 569 
20141642. 570 
Arnold, P. 2021. Evolution of the mammalian neck from developmental, morpho-functional, 571 
and paleontological perspectives. Journal of Mammalian Evolution, 28, 173-183. 572 
Arnold, P., Amson, E. and Fischer, M.S. 2017. Differential scaling patterns of vertebrae and 573 
the evolution of neck length in mammals. Evolution, 71, 1587-1599. 574 
24 
 
Badyaev, A.V., Foresman, K.R. and Young, R.L. 2005. Evolution of morphological 575 
integration: developmental accommodation of stress-induced variation. The American 576 
Naturalist, 166, 382-395. 577 
Bazzana, K. D., Evans, D. C., Bevitt, J. J. & Reisz, R. R. (2021). Neurosensory anatomy of 578 
Varanopidae and its implications for early synapsid evolution. Journal of Anatomy, In 579 
Press 580 
Benoit, J., Ford, D. P., Miyamae, J. A., & Ruf, I. (2021). Can maxillary canal morphology 581 
inform phylogenetic affinities? Acta Palaeontologica Polonica, 66, 389-393 582 
Benson, R. B. (2012). Interrelationships of basal synapsids: cranial and postcranial 583 
morphological partitions suggest different topologies. Journal of Systematic 584 
Palaeontology, 10, 601-624. 585 
Benson, R. B., Hunt, G., Carrano, M. T., & Campione, N. (2018). Cope’s rule and the 586 
adaptive landscape of dinosaur body size evolution. Palaeontology, 61, 13-48 587 
Berman, D. S., Maddin, H. C., Henrici, A. C., Sumida, S. S., Scott, D., & Reisz, R. R. (2020). 588 
New primitive caseid (Synapsida, Caseasauria) from the early Permian of Germany. 589 
Annals of the Carnegie Museum, 86, 43-75 590 
Biewener, A. A. (1982). Bone strength in small mammals and bipedal birds: do safety factors 591 
change with body size? Journal of Experimental Biology, 98, 289-301. 592 
Brochu, C. (1997). Morphology, fossils, divergence timing and the phylogenetic relationships 593 
of Gavialis. Systematic Biology, 46, 479-522 594 
Brocklehurst, N. (2016). Rates and modes of body size evolution in early carnivores and 595 
herbivores: a case study from Captorhinidae. PeerJ, 4, e1555. 596 
Brocklehurst, N. (2017). Rates of morphological evolution in Captorhinidae: an adaptive 597 
radiation of Permian herbivores. PeerJ, 5, e3200. 598 
25 
 
Brocklehurst, N. (2020). Olson’s Gap or Olson’s Extinction? A Bayesian tip-dating approach 599 
to resolving stratigraphic uncertainty. Proceedings of the Royal Society B, 287, 600 
20200154 601 
Brocklehurst, N. (2021). The First Age of Reptiles? Comparing reptile and synapsid 602 
diversity, and the influence of Lagerstätten, during the Carboniferous and early 603 
Permian. Frontiers in Ecology and Evolution, 9, 669765.  604 
Brocklehurst, N., & Brink, K. S. (2017). Selection towards larger body size in both 605 
herbivorous and carnivorous synapsids during the Carboniferous. Facets, 2, 68-84. 606 
Brocklehurst, N., & Fröbisch, J. (2018). A reexamination of Milosaurus mccordi, and the 607 
evolution of large body size in Carboniferous synapsids. Journal of Vertebrate 608 
Paleontology, 38, e1508026. 609 
Brocklehurst, N., & Benson, R. J. B. (2021). Multiple paths to morphological diversification 610 
during the origin of amniotes. Nature Ecology and evolution, 5, 1243-1249 611 
Brocklehurst, N., Reisz, R. R., Fernandez, V., Fröbisch, J. (2016).  A Re-Description of 612 
‘Mycterosaurus’ smithae, an Early Permian Eothyridid, and Its Impact on the 613 
Phylogeny of Pelycosaurian-Grade Synapsids. PLoS ONE, 11, e0156810 614 
Brocklehurst, N., Kammerer, C. F., & Benson, R. J. (2020). The origin of tetrapod herbivory: 615 
effects on local plant diversity. Proceedings of the Royal Society B, 287, 20200124. 616 
Brocklehurst, N., Panciroli, E., Benevento, G. L., & Benson, R. B. (2021). Mammaliaform 617 
extinctions as a driver of the morphological radiation of Cenozoic mammals. Current 618 
Biology, 31, 2955-2963 619 
Campione, N. E., & Evans, D. C. (2012). A universal scaling relationship between body mass 620 
and proximal limb bone dimensions in quadrupedal terrestrial tetrapods. BMC 621 
Biology, 10, 1-22 622 
26 
 
Carroll R. L. (1970) Quantitative aspects of the amphibian-reptilian transition. Forma et 623 
Functio, 3, 165-178 624 
Clauss, M., & Hummel, J. (2005). The digestive performance of mammalian herbivores: why 625 
big may not be that much better. Mammal Review, 35, 174-187 626 
Compton, A. W., & Jenkins, F. A. (1968). Molar occlusion in Late Triassic mammals. 627 
Biological Reviews, 43, 427-458 628 
Ezcurra, M. D., & Butler, R. J. (2018). The rise of the ruling reptiles and ecosystem recovery 629 
from the Permo-Triassic mass extinction. Proceedings of the Royal Society B, 285, 630 
20180361. 631 
Fabre, A.C., Perry, J.M., Hartstone‐Rose, A., Lowie, A., Boens, A. and Dumont, M., 2018. 632 
Do muscles constrain skull shape evolution in Strepsirrhines? The Anatomical 633 
Record, 301, pp.291-310. 634 
Felice, R.N. and Goswami, A., 2018. Developmental origins of mosaic evolution in the avian 635 
cranium. Proceedings of the National Academy of Sciences, 115, 555-560.  636 
Felice, R.N., Watanabe, A., Cuff, A.R., Noirault, E., Pol, D., Witmer, L.M., Norell, M.A., 637 
O'Connor, P.M. and Goswami, A. 2019. Evolutionary integration and modularity in 638 
the archosaur cranium. Integrative and comparative biology, 59, 371-382. 639 
Foote, M. (1994). Morphological disparity in Ordovician-Devonian crinoids and the early 640 
saturation of morphological space. Paleobiology, 20, 320-344 641 
Ford, D. P., & Benson, R. B. (2019). A redescription of Orovenator mayorum (Sauropsida, 642 
Diapsida) using high‐resolution μ CT, and the consequences for early amniote 643 
phylogeny. Papers in Palaeontology, 5, 197-239. 644 
Ford, D. P., & Benson, R. B. (2020). The phylogeny of early amniotes and the affinities of 645 
Parareptilia and Varanopidae. Nature Ecology & Evolution, 4, 57-65. 646 
27 
 
Frazzetta, T. H. (1968). Adaptive problems and possibilities in the temporal fenestration of 647 
tetrapod skulls. Journal of Morphology, 125, 145-157. 648 
Fröbisch, J., Schoch, R.R., Müller, J., Schindler, T. and Schweiss, D., 2011. A new basal 649 
sphenacodontid synapsid from the Late Carboniferous of the Saar-Nahe Basin, 650 
Germany. Acta Palaeontologica Polonica, 56, 113-120. 651 
Grunert, H. R., Brocklehurst, N., & Fröbisch, J. (2019). Diversity and disparity of 652 
Therocephalia: Macroevolutionary patterns through two mass extinctions. Scientific 653 
reports, 9(1), 1-11. 654 
Haridy, Y., Macdougall, M. J., Scott, D., & Reisz, R. R. (2016). Ontogenetic change in the 655 
temporal region of the Early Permian parareptile Delorhynchus cifellii and the 656 
implications for closure of the temporal fenestra in amniotes. PloS one, 11, e0166819. 657 
Heath, T. A., Huelsenbeck, J. P., & Stadler, T. (2014). The fossilized birth–death process for 658 
coherent calibration of divergence-time estimates. Proceedings of the National 659 
Academy of Sciences, 111, E2957-E2966. 660 
Hotton, N.I.I.I., Olson, E.C., Beerbower, R., Sumida, S.S. and Martin, K.L.M. 1997. Amniote 661 
origins and the discovery of herbivory. In Amniote origins (pp. 207-264). Academic 662 
Press. 663 
Janis, C. M., & Keller, J. C. (2001). Modes of ventilation in early tetrapods: Costal aspiration 664 
as a key feature of amniotes. Acta Palaeontologica Polonica, 46, 137-170. 665 
Jenkins Jr, F. A. (1974). The movement of the shoulder in claviculate and aclaviculate 666 
mammals. Journal of Morphology, 144, 71-83. 667 
Jenkins, Jr, F. A., & Parrington, F. R. (1976). The postcranial skeletons of the Triassic 668 
mammals Eozostrodon, Megazostrodon and Erythrotherium. Philosophical 669 
Transactions of the Royal Society of London. B, 273, 387-431. 670 
28 
 
Jombart, T., Balloux, F., & Dray, S. (2010). Adephylo: new tools for investigating the 671 
phylogenetic signal in biological traits. Bioinformatics, 26, 1907-1909. 672 
Jones, K. E., Angielczyk, K. D., Polly, P. D., Head, J. J., Fernandez, V., Lungmus, J. K., 673 
Tulga, S. & Pierce, S. E. (2018). Fossils reveal the complex evolutionary history of 674 
the mammalian regionalized spine. Science, 361, 1249-1252. 675 
Jones, K. E., Gonzalez, S., Angielczyk, K. D., & Pierce, S. E. (2020). Regionalization of the 676 
axial skeleton predates functional adaptation in the forerunners of mammals. Nature 677 
ecology & evolution, 4, 470-478. 678 
Jones, K. E., Dickson, B. V., Angielczyk, K. D., & Pierce, S. E. (2021). Adaptive landscapes 679 
challenge the “lateral-to-sagittal” paradigm for mammalian vertebral evolution. 680 
Current Biology, 31, 1883-1892. 681 
Laurin, M. (2004). The evolution of body size, Cope's rule and the origin of amniotes. 682 
Systematic Biology, 53, 594-622. 683 
Lee, M.S. and Spencer, P.S., 1997. Crown-clades, key characters and taxonomic stability: 684 
when is an amniote not an amniote? In Amniote origins (pp. 61-84). Academic Press. 685 
Liu, J., Rubidge, B. S. & Li, J. (2009). New basal synapsid supports Laurasian origin for 686 
therapsids. Acta Palaeontologica Polonica, 54, 393-400 687 
Lloyd, G. T. (2016). Estimating morphological diversity and tempo with discrete character-688 
taxon matrices: implementation, challenges, progress, and future directions. 689 
Biological Journal of the Linnean Society, 118, 131-151. 690 
Lucas, S. G., & Heckert, A. B. (2001) Olson’s gap: a global hiatus in the record of middle 691 
Permian tetrapods. Journal of Vertebrate Paleontology 21, 75A 692 
Lungmus, J. K., & Angielczyk, K. D. (2021). Phylogeny, function and ecology in the deep 693 
evolutionary history of the mammalian forelimb. Proceedings of the Royal Society B, 694 
288, 20210494. 695 
29 
 
Luo, Z. X., Gatesy, S. M., Jenkins, F. A., Amaral, W. W. & Shubin, N. H. (2015). 696 
Mandibular and dental characteristics of Late Triassic mammaliaform Haramiyavia 697 
and their ramifications for basal mammal evolution. Proceedings of the National 698 
Academy of Sciences, 112, E7101-E7109 699 
MacDougall, M. J., & Reisz, R. R. (2014). The first record of a nyctiphruretid parareptile 700 
from the Early Permian of North America, with a discussion of parareptilian temporal 701 
fenestration. Zoological Journal of the Linnean Society, 172(3), 616-630. 702 
MacDougall, M. J., Brocklehurst, N., & Fröbisch, J. (2019). Species richness and disparity of 703 
parareptiles across the end-Permian mass extinction. Proceedings of the Royal Society 704 
B, 286, 20182572. 705 
Maddin, H. C., Mann, A., Herbert, B. (2020). Varanopid from the Carboniferous of Nova 706 
Scotia reveals the evidence of parental care in amniotes. Nature Ecology and 707 
Evolution, 4, 50-56 708 
Mann, A., Dudgeon, T.W., Henrici, A.C., Berman, D.S. and Pierce, S.E., 2021. Digit and 709 
ungual morphology suggest adaptations for scansoriality in the late Carboniferous 710 
eureptile Anthracodromeus longipes. Frontiers in Earth Science, 9, 440. 711 
Modesto S. P. (1992). Did herbivory foster early amniote diversification? Journal of 712 
Vertebrate Paleontology, 12 (Suppl.), 44A 713 
Modesto, S. P., Lamb, A. J., & Reisz, R. R. (2014). The captorhinid reptile Captorhinikos 714 
valensis from the lower Permian Vale Formation of Texas, and the evolution of 715 
herbivory in eureptiles. Journal of Vertebrate Paleontology, 34, 291-302. 716 
Modesto, S. P., Scott, D. M., MacDougall, M. J., Sues, H. D., Evans, D. C., & Reisz, R. R. 717 
(2015). The oldest parareptile and the early diversification of reptiles. Proceedings of 718 
the Royal Society B: Biological Sciences, 282, 20141912. 719 
30 
 
Müller, D. W., Codron, D., Meloro, C., Munn, A., Schwarm, A., Hummel, J., & Clauss, M. 720 
(2013). Assessing the Jarman-Bell principle: scaling of intake, digestibility, retention 721 
time and gut fill with body mass in mammalian herbivores. Comparative 722 
Biochemistry and Physiology A, 164, 129-140 723 
Payton, M. E., Greenstone, M. H., & Schenker, N. (2003). Overlapping confidence intervals 724 
or standard error intervals: what do they mean in terms of statistical significance? 725 
Journal of Insect Science, 3, 34. 726 
Piñeiro, G., Ferigolo, J., Ramos, A., & Laurin, M. (2012). Cranial morphology of the Early 727 
Permian mesosaurid Mesosaurus tenuidens and the evolution of the lower temporal 728 
fenestration reassessed. Comptes Rendus Palevol, 11, 379-391. 729 
 R Core Team (2019). R: A language and environment for statistical computing. R 730 
Foundation for Statistical Computing, Vienna, Austria. 731 
Reisz, R. R., & Fröbisch, J. (2014). The oldest caseid synapsid from the Late Pennsylvanian 732 
of Kansas, and the evolution of herbivory in terrestrial vertebrates. PloS one, 9, 733 
e94518. 734 
Revell, L. J. (2012). phytools: an R package for phylogenetic comparative biology (and other 735 
things). Methods in Ecology and Evolution, 3, 217-223. 736 
Rhoda, D., Polly, P.D., Raxworthy, C. and Segall, M. 2021. Morphological integration and 737 
modularity in the hyperkinetic feeding system of aquatic‐foraging snakes. Evolution, 738 
75, 56-72. 739 
Romano, M., Brocklehurst, N., & Fröbisch, J. (2017). Discrete and continuous character-740 
based disparity analyses converge to the same macroevolutionary signal: a case study 741 
from captorhinids. Scientific reports, 7, 1-9. 742 
Romer A. S. (1957) Origin of the amniote egg. The Scientific Monthly, 85, 57-63  743 
31 
 
Ronquist, F., & Huelsenbeck, J. P. (2003). MrBayes 3: Bayesian phylogenetic inference 744 
under mixed models. Bioinformatics, 19, 1572-1574. 745 
Rooney, N., & McCann, K. S. (2012). Integrating food web diversity, structure and stability. 746 
Trends in Ecology and Evolution, 27, 40-46 747 
Rubidge, B. S. & Sidor, C. A. (2001). Evolutionary patterns among Permo-Triassic 748 
therapsids. Annual Review of Ecology and Systematics, 32, 449-480 749 
Runestad, J. A., & Ruff, C. B. (1995). Structural adaptations for gliding in mammals with 750 
implications for locomotor behavior in paromomyids. American Journal of Physical 751 
Anthropology, 98, 101-119. 752 
Ruta, M., Wagner, P. J., & Coates, M. I. (2006). Evolutionary patterns in early tetrapods. I. 753 
Rapid initial diversification followed by decrease in rates of character change. 754 
Proceedings of the Royal Society B, 273, 2107-2111. 755 
Ruta, M., Angielczyk, K. D., Fröbisch, J., & Benton, M. J. (2013a). Decoupling of 756 
morphological disparity and taxic diversity during the adaptive radiation of 757 
anomodont therapsids. Proceedings of the Royal Society B: Biological Sciences, 280, 758 
20131071. 759 
Ruta, M., Botha-Brink, J., Mitchell, S. A., & Benton, M. J. (2013b). The radiation of 760 
cynodonts and the ground plan of mammalian morphological diversity. Proceedings 761 
of the Royal Society B: Biological Sciences, 280, 20131865. 762 
Ruta, M., Krieger, J., Angielczyk, K. D., & Wills, M. A. (2018). The evolution of the 763 
tetrapod humerus: Morphometrics, disparity, and evolutionary rates. Earth and 764 
Environmental Science Transactions of the Royal Society of Edinburgh, 109, 351-765 
369. 766 
Simpson, G. G. (1953). The major features of evolution. Columbia University Press. 767 
32 
 
Spindler, F., Werneburg, R., Schneider, J. W., Luthardt, L., Annacker, V., Rößler, R. (2018). 768 
First arboreal ‘pelycosaurs’ (Synapsida: Varanopidae) from the early Permian 769 
Chemnitz fossil lagerstätte, SE Germany, with a review of varanopid phylogeny. 770 
Paläontologische Zeitschrift, 92, 315-364 771 
Stanley, S.M., 1973. An explanation for Cope's rule. Evolution, 27, 1-26. 772 
Sues, H.D. and Reisz, R.R., 1998. Origins and early evolution of herbivory in tetrapods. 773 
Trends in Ecology & Evolution, 13, 141-145. 774 
Watanabe, A., Fabre, A.C., Felice, R.N., Maisano, J.A., Müller, J., Herrel, A. and Goswami, 775 
A. 2019. Ecomorphological diversification in squamates from conserved pattern of 776 
cranial integration. Proceedings of the National Academy of Sciences, 116, 14688-777 
14697. 778 
Weaver, L. N., & Grossnickle, D. M. (2020). Functional diversity of small-mammal 779 
postcrania is linked to both substrate preference and body size. Current zoology, 66, 780 
539-553. 781 
Werneburg, I. (2019). Morphofunctional categories and ontogenetic origin of temporal skull 782 
openings in amniotes. Frontiers in Earth Science, 7, 13. 783 
Woodward, G., Ebenman, B., Emmerson, M., Montoya, J. M., Olesen, J. M., Valido, A., 784 
Warren, P. H. (2005). Body size in ecological networks. Trends in Ecology and 785 
Evolution, 20, 402-409 786 
Young, R.L. and Badyaev, A.V. 2006. Evolutionary persistence of phenotypic integration: 787 
influence of developmental and functional relationships on complex trait evolution. 788 
Evolution, 60, 1291-1299. 789 
 790 
 791 
Figure Captions 792 
33 
 
Figure 1: The maximum clade credibility tree produced from the Fossilised Birth Death 793 
analysis, with branches colour coded according to log transformed evolutionary rates. Nodes 794 
discussed in the text labelled: Eup. – Eupelycosauria; Var. – Varanopidae; Par. – Parareptilia; 795 
Sph. – Sphenacodontia; Cas. – Caseidae; Ther. – Therapsida; Neo. – Neodiapsida 796 
 797 
Figure 2: Rates of evolution through time. Narrow lines represent rates of individual 798 
branches. Mid Weight line represents the median rate of each 1 million year time slice. Thick 799 
line represents loess fitted regression between median rate and time. A) All amiotes; B) 800 
reptiles and synapsids compared 801 
 802 
Figure 3: Patterns of constraint in amniotes. A) Significant variation in constraint plotted over 803 
a phylogeny where branch lengths represent the amount of character change along the branch. 804 
Nodes in red experience a significant relaxation of constraint. Nodes in blue experience a 805 
significant strengthening of constraints. B) Comparison of Patristic distances and 806 
morphological dissimilarity in Synapsids (red) and Reptiles (blue). Each point represents a 807 
pairwise comparison of two taxa. The curves represent Loess fitted regression curves. 808 
 809 
Figure 4: Patterns of constraint in amniotes within different anatomical partitions plotted over 810 
phlogenies where branch lengths represent the amount of character change within the 811 
character prtition along that branch. Nodes in red experience a significant relaxation of 812 
constraint. Nodes in blue experience a significant strengthening of constraints. A) Skull; B) 813 
Snout; C) Temporal region; D) Postcranium. 814 
 815 
Figure 5: Rates of evolution in reptiles and synapsids within anatomical partotions through 816 
time. Narrow lines represent rates of individual branches. Mid Weight line represents the 817 
34 
 
median rate of each 1 million year time slice. Thick line represents loess fitted regression 818 
between median rate and time. A) Skull; B) Postcranium 819 
 820 
 821 
Figure 6: A) Body size category assigned to each taxon, and likelihood ancestral state 822 
reconstruction of body size categories over the time calibrated tree. Colour of tip represents 823 
size assigned to that tip. Pie charts at each node represent relative probability of each size 824 
category being the ancestral state of that node. B) Patterns of constraint within each size 825 
category. Each point represents a pairwise comparison of two taxa within a size category. The 826 
curves represent Loess fitted regression curve. C) Patterns of constraint within reptiles and 827 
synapsids assigned to the large or very large categories. 828 
 829 
Figure 7: Rates of body size evolution in Synapsids and Reptiles. Numbers alongside arrows 830 
represent the rate of transition between the size categories (the instantaneous probability of 831 
transition) in the direction indicated by the arrow. Transitions with a rate of 0 are not shown. 832 
The colour of the arrow represents the log transformed rates. Silhouettes open source from 833 
phylopic.org (not to scale) 834 
 835