stubbs, t. l. , & benton, m. j. (2016). ecomorphological ... · thomas l. stubbs and michael j....

Stubbs, T. L., & Benton, M. J. (2016). Ecomorphologicaldiversifications of Mesozoic marine reptiles: The roles of ecologicalopportunity and extinction. Paleobiology, 42(4), 547-573.https://doi.org/10.1017/pab.2016.15

Publisher's PDF, also known as Version of recordLicense (if available):CC BYLink to published version (if available):10.1017/pab.2016.15

Link to publication record in Explore Bristol ResearchPDF-document

This is the final published version of the article (version of record). It first appeared online via CambridgeUniversity Press atjournals.cambridge.org/action/displayAbstract?fromPage=online&aid=10322544&fileId=S0094837316000154

University of Bristol - Explore Bristol ResearchGeneral rights

This document is made available in accordance with publisher policies. Please cite only thepublished version using the reference above. Full terms of use are available:http://www.bristol.ac.uk/pure/user-guides/explore-bristol-research/ebr-terms/

https://doi.org/10.1017/pab.2016.15

https://doi.org/10.1017/pab.2016.15

https://research-information.bris.ac.uk/en/publications/c3f9b0fe-6bc2-400c-a770-1d1ab6e8379e

https://research-information.bris.ac.uk/en/publications/c3f9b0fe-6bc2-400c-a770-1d1ab6e8379e

Paleobiologyhttp://journals.cambridge.org/PAB

Additional services for Paleobiology:

Email alerts: Click hereSubscriptions: Click hereCommercial reprints: Click hereTerms of use : Click here

Ecomorphological diversications of Mesozoic marine reptiles: the roles ofecological opportunity and extinction

Thomas L. Stubbs and Michael J. Benton

Paleobiology / FirstView Article / May 2016, pp 1 - 27DOI: 10.1017/pab.2016.15, Published online: 17 May 2016

Link to this article: http://journals.cambridge.org/abstract_S0094837316000154

How to cite this article:Thomas L. Stubbs and Michael J. Benton Ecomorphological diversications of Mesozoic marinereptiles: the roles of ecological opportunity and extinction. Paleobiology, Available on CJO 2016doi:10.1017/pab.2016.15

Request Permissions : Click here

Downloaded from http://journals.cambridge.org/PAB, IP address: 137.222.120.226 on 23 May 2016

Ecomorphological diversifications of Mesozoic marine reptiles:the roles of ecological opportunity and extinction

Thomas L. Stubbs and Michael J. Benton

Abstract.—Mesozoic marine ecosystems were dominated by several clades of reptiles, includingsauropterygians, ichthyosaurs, crocodylomorphs, turtles, and mosasaurs, that repeatedly invaded oceanecosystems. Previous research has shown that marine reptiles achieved great taxonomic diversity in theMiddle Triassic, as they broadly diversified into many feeding modes in the aftermath of the Permo-Triassicmass extinction, but it is not known whether this initial phase of evolution was exceptional in the context ofthe entire Mesozoic. Here, we use a broad array of disparity, morphospace, and comparative phylogeneticanalyses to test this. Metrics of ecomorphology, including functional disparity in the jaws and dentition andskull-size diversity, show that theMiddle to early Late Triassic represented a time of pronounced phenotypicdiversification in marine reptile evolution. Following the Late Triassic extinctions, diversity recovered, butdisparity did not, and it took over 100Myr for comparable variation to recover in the Campanian andMaastrichtian. Jurassicmarine reptiles generally failed to radiate into vacated functional roles. The signaturesof adaptive radiation are not seen in all marine reptile groups. Clades that diversified during the Triassicbiotic recovery, the sauropterygians and ichthyosauromorphs, do show early diversifications, early highdisparity, and early burst, while less support for these models is found in thalattosuchian crocodylomorphsandmosasaurs. Overall, the Triassic represented a special interval inmarine reptile evolution, as a number ofgroups radiated into new adaptive zones.

Thomas L. Stubbs and Michael J. Benton. School of Earth Sciences, University of Bristol, Bristol BS8 1RJ, U.K.E-mail: [email protected]

Accepted: 14 March 2016Supplemental materials deposited at Dryad: doi:10.5061/dryad.5890r

IntroductionThe Mesozoic Era witnessed the rise of

tetrapods as dominant components in marineecosystems (Kelley and Pyenson 2015). Theemergence of a diverse assemblage of marinereptiles in the Triassic marked the establishmentof ecosystem complexity on a par with modernoceans (Fröbisch et al. 2013; Liu et al. 2014).During faunal recovery following the devastat-ing Permo-Triassic mass extinction (PTME),sauropterygians, ichthyosaurs, and thalatto-saurs diversified within the marine realm,creating and filling trophic niches that had notbeen widely exploited in the Paleozoic (Bentonet al. 2013). Although most marine reptilelineages disappeared in the Late Triassic,secondary Jurassic and Cretaceous radiationsoccurred among the surviving plesiosaurs andparvipelvian ichthyosaurs (Motani 2005; Bensonet al. 2012). Other independent invasions of theoceans by thalattosuchian crocodylomorphs,marine turtles, and mosasaurs in the Jurassicand Cretaceous meant that reptiles remained at

the apex of marine ecosystems throughout theentire Mesozoic (Young et al. 2010; Benson andButler 2011; Polcyn et al. 2014).

The PTME devastated preexisting biotas, andthe subsequent Triassic biotic recovery has beenrecognized as an exceptional period in thediversification of life (Chen and Benton 2012).The rise of marine reptiles as an adaptiveassemblage is considered one of the mostsignificant components of this recovery (Liuet al. 2014;Motani et al. 2015a).Numerical studiesshow that marine reptile species diversity firstpeaked in the Anisian, just 10Myr into theTriassic (Benson et al. 2010; Benson and Butler2011). In addition, qualitative observations fromthe fossil record reveal that marine reptilesbroadly diversified into a variety of trophicstrategies in their first 20Myr of evolution(Motani et al. 2015a). These included large-skulled macropredatory ichthyosaurs (Fröbischet al. 2013), small suction-feeding edentulousichthyosauromorphs (Motani et al. 2015b),shell-crushing placodonts (Neenan et al. 2014),

Paleobiology, page 1 of 27DOI: 10.1017/pab.2016.15

© 2016 The Paleontological Society. All rights reserved. This is anOpenAccess article, distributed under the terms of the CreativeCommons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-se, distribution,and reproduction in any medium, provided the original work is properly cited. 0094-8373/16

doi:10.5061/dryad.5890r

long-necked and fanged nothosaurs (Rieppel2002), bizarre heterodont thalattosaurs (Bentonet al. 2013), and enigmatic apparent filter-feeders,such as Atopodentatus (Cheng et al. 2014).These disparate dietary habits were associatedwith an exceptional array of ecomorphologicalspecializations, most linked to feeding andprey acquisition.

Recent research has emphasized that abroad-based ecological approach is essentialfor understanding marine reptile evolution(Motani et al. 2015a). Thus far, studies havefocused on body-size trends and trophic webestablishment in the first 5–10Myr of theMesozoic (Liu et al. 2014; Scheyer et al. 2014),while others have examined the proportions ofcategorical ecomorphs and guilds through thewhole Triassic (Benton et al. 2013; Kelley et al.2014; Motani et al. 2015a). However, no studyhas examined whether the Triassic was truly atime of unusual trophic proliferation orwhether the accumulated diversity of formswas any greater than at other points in theMesozoic –was the first wave of marine reptileevolution in the aftermath of the PTME reallyexceptional? A robust quantitative approach isrequired to test this, and trends of marinereptile ecomorphological diversity through thewhole Mesozoic must be quantified. In thisstudy, we focus on two elements of phenotypicdiversity. We quantify patterns of functionaldisparity in the jaws and dentition, focusingon the diversity of forms associated withfeeding and prey acquisition (Anderson 2009;Anderson et al. 2011; Stubbs et al. 2013), andseparately examine the diversity of skull sizes.

The transitions from land to sea in Mesozoicmarine reptiles represent excellent casestudies for exploring the macroevolutionaryconsequences of ecological opportunity.Invasion of new habitat opens up previouslyunexplored niche space and is considered oneof the major ecological opportunities in nature(Schluter 2000). Such opportunities are theprimary catalysts for bursts of phenotypicevolution and adaptive radiation (Simpson1953; Yoder et al. 2010). Mesozoic marinereptiles are a polyphyletic assemblage, madeup of species from independent evolutionaryinvasions of the marine realm from differentparts of the neodiapsid tree (Bardet et al. 2014).

These individual diversifications provide achance to contrast patterns of evolutionassociated with the ascent of several distinctclades in response to the same ecologicalopportunity. As ecological opportunity drivesadaptive radiation, one could hypothesizethat niche filling or “early-burst” diversifica-tion patterns should be universal in marinereptiles. However, there are a number ofconfounding factors. For example, not allmarine reptiles diversified simultaneously.Do clades originating in the immediateaftermath of major extinction events showsimilar trajectories and rates of evolution asthose diversifying at times when there were nomajor biotic perturbations?

There are two commonly applied quantita-tive methods to identify the signatures ofadaptive radiation in deep-time data. Mostattention has focused on examining time-seriestrends of lineage diversity and morphologicaldisparity (e.g., Erwin 2007; Ruta et al. 2013;Hughes et al. 2013). When maximumnumerical and morphological diversity is seenearly in a clade’s history, this indicates earlydiversification and morphospace expansionand can be used to infer an adaptive radiation(Benton et al. 2014). Recently, focus hasswitched to explicitly modeling evolutionaryrates in ecologically relevant traits in aphylogenetic framework (e.g., Benson et al.2014). Under a model of adaptive radiation,rapid evolutionary rates are seen early in aclade’s history, and these decelerate throughtime, a trend formalized as the “early-burst”model (Harmon et al. 2010). Here, we use acomparative approach to explore trends oflineage diversity, functional disparity,skull-size diversity, and rates of evolution inthe most diverse marine reptile clades to testthe prediction of universal early high diversityand disparity and early burst.

Two major turnover events are believed tohave significantly impacted Mesozoic marinereptile macroevolution. The first occurredduring the Late Triassic, when species diversitycrashed and whole clades and morphotypesdisappeared, including placodont sauroptery-gians, nothosaurs, thalattosaurs, and all non-parvipelvian ichthyosaurs (Benson et al. 2010;Benson and Butler 2011; Thorne et al. 2011;

2 THOMAS L. STUBBS AND MICHAEL J. BENTON

Kelley et al. 2014). A second, less severeextinction interval has been posited for theJurassic/Cretaceous boundary (Bardet 1994),although recent analyses show weak supportfor an exceptional loss of diversity, but pointrather to a “thinning out” of major clades,such as thalattosuchians, pliosaurids, andcryptoclidid plesiosauroids (Benson andDruckenmiller 2014). While these proposedextinction intervals are defined by losses oftaxonomic diversity, they could also have hadprofound impacts on ranges of morphology andecology exhibited by surviving taxa (Bapst et al.2012; Ruta et al. 2013).Extinctions can be nonrandom and highly

selective, with particular clades, ecologies,morphotypes, or adaptive zones beingsusceptible to biotic perturbations (Friedman2009; Korn et al. 2013). By dividing Triassic andEarly Jurassic marine reptiles into broadecotypes, Kelley et al. (2014) demonstratedthat shallow-marine durophages were mostsusceptible to sea-level fluctuations during theLate Triassic, while pelagic forms weremore resilient and able to avoid widespreadextinction. Quantitative studies have focusedon individual marine reptile clades. Forexample, Thorne et al. (2011) revealed amassive loss of ichthyosaur disparity duringthe Late Triassic/Early Jurassic transition, andshowed that surviving taxa occupieddifferent morphospace than Triassic forms. Byexpanding on this numerical approach andconsidering marine reptiles as an adaptiveassemblage, we assess what ecomorphologicaltraits are characteristic of extinction victimsthrough the Mesozoic and test for evidence ofinfilling in vacant adaptive zones by differentmarine reptile groups.

Materials and Methods

Taxon Sampling.—This study representsthe largest comparative and quantitativeinvestigation of phenotypic evolution inMesozoic marine reptiles. In total, 206 marinereptile species are used to investigate trends offunctional disparity in the jaws and dentition(Supplementary Data), and 354 species areincluded in the analyses of skull-size trends

(Supplementary Data). Individual specimensrepresent each species, and taxa range in agefrom the Olenekian to the end Maastrichtian.Five monophyletic clades are included:the sauropterygians, ichthyosauromorphs,thalattosaurs, thalattosuchian crocodylomorphs,and mosasauroids. Additionally, we samplemarine representatives from clade Testudinata,incorporating the stem turtle Odontochelys;Jurassic plesiochelyids; and the Cretaceous cladeChelonioidea. In the primary data set we alsoinclude a number of smaller clades and individualgenera known to have inhabitedMesozoicmarineenvironments: tanystropheids, saurosphargids,Helveticosaurus and Qianosuchus from theTriassic, and nonmosasauroid squamates fromthe Cretaceous. Late Jurassic pleurosaurids andCretaceous marine snakes, dyrosaurids, andpholidosaurids were not included in the datasets, due to a lack of material. When testingfor adaptive radiation based on time-seriestrends and evolutionary rates, investigation isfocused on the five most diverse clades:Sauropterygia, Eosauropterygia (Sauropterygiaminus Placodontia), Ichthyosauromorpha,Thalattosuchia, and Mosasauroidea. Separateanalyses were performed for Eosauropterygia toaccount for the aberrant morphology ofplacodonts. Each of these clades could beeffectively sampled on an individual basis interms of both phylogenetic and strati-graphic coverage. The thalattosaurs, turtles,tanystropheids, and saurosphargids could not beincorporated into these comparisons because ofsmaller sample sizes and sporadic stratigraphicoccurrences.

Functional Disparity.—When examiningfunctional disparity, focus is placed onquantifying the diversity of forms andinnovations that have known ecomorphologicaland/or biomechanical consequences and aretherefore directly associated with resource useand the acquisition and processing of prey(e.g., Anderson et al. 2011, 2013; Stubbs et al.2013; Button et al. 2014). Functional and/orbiomechanical traits are primarily derived fromthe mandible and dentition. Figure 1 illustrates arange of marine reptile jaws, highlighting thegreat diversity of forms.

Here, functional disparity was assessedusing a combination of continuous measured

MESOZOIC MARINE REPTILE DISPARITY 3

traits and discrete characters (Table 1). Thenine continuous characters were measured for206 jaws in the program ImageJ (Schneideret al. 2012). For consistency, all measurementswere recorded digitally from photographscollected during museum visits, photographsprovided by colleagues, figures sourcedfrom published literature, and anatomical

reconstructions. By recording all linearmeasurements digitally from two-dimensionalimages, we aimed to avoid potentialdiscrepancies and distortions that could arisethrough collecting some data from directmeasurements of three-dimensional objectsand other data from two-dimensional images.

All but one of the nine discrete charactersrelate to dental features, which are oftendifficult to quantify with continuous variablesbut have considerable functional implications(Massare 1987). Detailed descriptions of allcharacters are provided in the SupplementaryText, and raw data are provided in theSupplementary Data. The continuous functionalcharacters (C1–9) were standardized using thez-transformation (so the mean charactervalue was 0, with a standard deviation of 1)(Anderson et al. 2011). Separate standardizationswere usedwhen analyzing the complete data set,incorporating all marine reptiles and subsets ofthe data for different clades.

Functional Morphospaces.—Functional mor-phospaces were constructed using multi-variate ordination techniques. The full data set,composed of z-transformed continuous variablesand binary scores for all 206 marine reptilespecies, was used to calculate an intertaxondistance matrix based on Gower’s similaritycoefficient (Gower 1971). This similaritymetric was selected because it does nothave metric properties and can therefore becomputed with mixed data types (continuousand discrete). The similarity matrix wasanalyzed using principal coordinates analysis(PCOa) to produce multivariate ordination axessummarizing major elements of phenotypic

FIGURE 1. A sample of jaw functional morphotypes fromthe fossil record of Mesozoic marine reptiles. The illustratedtaxa are (A) Pliosaurus, (B) Tylosaurus, (C) Ophthalmosaurus,(D) Metriorhynchus, (E) Nothosaurus, (F), Xinpusaurus,(G) Placochelys, and (H) Mesodermochelys. Scale bars on thejaw illustrations represent 20 cm (A–D) and 5cm (E–H).

TABLE 1. Functional characters measured and scored for 206 Mesozoic marine reptile taxa. Detailed descriptions of allcharacters are provided in the Supplementary Text.

Functional characters

Continuous traits Discrete characters

C1: Maximum jaw depth/length C10: Intramandibular joint: absent/presentC2: Average jaw depth/length C11: Pointed and recurved tooth crowns: absent/presentC3: Relative length of the mandibular symphysis C12: Anterior part of dentary: dentigerous/edentulousC4: Anterior mechanical advantage (AMA) C13: Enlarged procumbent dentary fangs: absent/presentC5: Posterior mechanical advantage (PMA) C14: Strongly procumbent chisel-shaped anterior dentary teeth:

absent/presentC6: Relative muscle attachment area C15: Fence-like dentition: absent/presentC7: Relative length of the retroarticular process C16: Bulbous crushing dentition: absent/presentC8: Relative length of the dental row C17: Posterior dentary tooth plates: absent/presentC9: Mandibular length C18: Cutting dentition: absent/present


variation. The scores (coordinates) of the taxa onPC axes 1, 2, and 3 were used to constructmorphospaces because these axes contribute themost to overall absolute variance. Only the first11 axes were retained for statistical tests anddisparity calculations, because higher axeswere associated with negligible amounts ofabsolute variance (<1%) or negative eigen-values (Table 2). To ensure that the presenceof negative eigenvalues did not distort theordinated dissimilarity among forms, weperformed correlation tests on the computedintertaxon distances and ordination distances(Friedman 2012). Results show strong correlationbetween the distance measures (Pearson’sproduct-moment correlation: r=0.982, p<0.001),suggesting the PCOa accurately preserves mostof the underlying distances between taxawithout major distortion (Supplementary Fig. 1).Nonparametric multivariate analyses of variance(NPMANOVA) were carried out on the 11retained PCOa axes to test for separation bytemporal and group bins in morphospace (Rutaet al. 2013). The above-described procedures wererepeated separately for each of the five majorclades to generate multivariate axes of variationfor subsequent disparity calculations in eachgroup (see below).Temporal Disparity Calculations.—Marine

reptile disparity was quantified using thePCOa scores of taxa from the first 11ordination axes (Table 2), based on the sum ofvariances and the sum of ranges metrics. Toderive a trajectory of disparity through time,we performed calculations on taxa binned in

16 Mesozoic time intervals. The sum of rangescalculations were rarefied to account fordiffering sample sizes, using both the averagesample size of all bins (n= 17) and the smallestsample size (n= 5). Statistical tests fordifferences in disparity between time bins arebased on 95% confidence intervals, paired-sample t-tests based on multivariate variance(Anderson et al. 2011), and ratios of marginallikelihoods for variance (Finarelli and Flynn2007). Confidence intervals associated withcalculated disparity values were generatedby bootstrapping with 10,000 replications.Ratios of marginal likelihoods were usedexclusively to identify shifts in disparitybetween successive time intervals based onthe procedure outlined in Finarelli and Flynn(2007) and Anderson et al. (2011). We assessedthe statistical dependence between bin samplesize and disparity by applying generalizeddifferencing and assessing the strength andsignificance of correlations (e.g., Ruta et al.2013). Partial disparity for each marinereptile group was also examined to seehow they contribute to overall disparitythrough the Mesozoic (Foote 1993). Alldisparity calculations and significancetests were performed using the MDAMatlab® package (Navarro 2003) and in R(R Development Core Team 2011).

Stratigraphic Binning.—For temporal disparitycalculations and morphospace plots, taxa werebinned in 16 Mesozoic time intervals. We used acombination of stage-level and composite bins,with ages in Myr: Olenekian (250–247.1), Anisian(247.1–241.5), Ladinian (241.5–237), Carnian(237–228.4), Norian (228.4–209.5), Hettangian–Sinemurian (201.3–190.8), Pliensbachian–Toarcian (190.8–174.1), Aalenian–Bathonian(174.1–166.1), Callovian–Oxfordian (166.1–157.3), Kimmeridgian–Tithonian (157.3–145),Berriasian–Barremian (145–126.3), Aptian–Albian (126.3–100.5), Cenomanian–Turonian(100.5–89.8), Coniacian–Santonian (89.77–83.6),Campanian (83.6–72.1), and Maastrichtian(72.1–66). A Rhaetian bin was not included dueto the lack of marine reptile fossils during thisinterval. Absolute ages for the stratigraphicstages are from Gradstein et al. (2012). Themean bin duration is 11Myr and the rangeis 22.9Myr.

TABLE 2. Eigenvalues and percentages of varianceassociated with each PCO axis from the multivariateanalysis. The primary analysis is based on all 18 characters.

Principal coordinate analysis

PC Eigenvalue % variance

1 2.6051 27.422 1.1496 12.103 1.0131 10.664 0.77706 8.185 0.32332 3.406 0.26086 2.757 0.15826 1.678 0.13675 1.449 0.13178 1.3910 0.11304 1.1911 0.094332 0.99


Stratigraphic assignments were based on theprimary literature, previous compendia ofmarine reptile diversity, and the PaleobiologyDatabase (paleodb.org) (Benson et al. 2010;Young et al. 2010; Benton et al. 2013; Kelley et al.2014). Species-level stratigraphic assignmentswere used in most cases, but in the disparityanalyses there were a number of instances inwhich species could not be sampled in a time bindue to inadequately preserved material. In thesecases, the stratigraphic range of a representativetaxon belonging to the same genuswas extendedto account for that taxon’s absence (Supplemen-tary Data). There were two instances in whichtaxa could not have their absence accounted forby a member of the same genus, both of them inthe Norian time bin. Because it is importantto represent all forms at a given time, wedecided to represent their ecomorphologicalcharacteristics with a closely related species;we used Psephochelys polyosteoderma to representthe placodont species Psephoderma alpinum(and placodonts in general), and we usedGuanlingsaurus liangae to represent the largeedentulous ichthyosaur Shonisaurus sikanniensis.

Disparity Sensitivity Analyses.—To scrutinizetemporal trends of marine reptile disparity weperformed sensitivity tests using alternativeprotocols and data subsets. First, disparity wasrecalculated based on within-bin mean pairwisedissimilarity from the original Gower intertaxondistance matrix for all 206 taxa (withoutPCOa ordination; e.g., Close et al. 2015).Second, disparity analyses were undertakenbased solely on the nine continuous functionalmetrics (C1–C9) measured across all taxa, toensure that the calculated trajectory of marinereptile disparity through time was notsimply the result of overwhelming dominancefrom binary characters. The ichthyosaurThalattoarchon was removed, because it couldnot be scored for enough continuous characters.Because size can dominate data sets based onlinear measurements and ratios, additionaltests were run on eight continuous variablesminus the character total mandibular length(C9). For both cases, the z-transformedcontinuous characters were converted into avariance-covariance matrix and subjected toprincipal components analysis (PCA) toderive multivariate ordination axes and the

corresponding PC scores for all 205 taxa.Missing values were accounted for usingiterative imputation with 10,000 bootstrapreplicates (Ilin and Raiko 2010). Sevenordination axes were retained from the nine-character analysis (accounting for 98.9% ofvariance), and six axes were retained from theanalysis with size removed (accounting for98.7% variance). Following the protocoldiscussed above, disparity based on the sum ofvariances and sum of ranges was calculated inthe 16 Mesozoic time intervals.

Skull-Size Trends.—To complement thedisparity analyses, we also examined patternsof skull-size evolution in all Mesozoic marinereptiles. Skull size represents an ecomorpho-logical characteristic that can be comparedbroadly across all marine reptile clades andused to identify patterns of phenotypicdiversity. In 354 marine reptile species,maximum skull length (MSL) was measuredfrom the anterior tip of the premaxilla to theposterior margin of the squamosal. Data werecollected during museum visits and frompublished tables and figures (SupplementaryData). For incomplete specimens, MSL wasestimated based on cranial proportions ofclosely related species or total mandiblelength (required for <5% of taxa). Juvenilespecimens, identified based on discussions inthe literature, were not included in the data set.Temporal trends of skull-size evolution wereexplored by plotting log10-transformed MSLagainst geological time based on thestratigraphic range midpoints of all species.Univariate disparity, based on the range andinterquartile range, was examined in thesame time bins as the multivariate disparityanalyses.

Comparative Disparity and DiversityAnalyses.—To test whether all marine reptileclades have similar trajectories of disparitythrough time, we compared the temporaldisparity profiles of sauropterygians, eosauro-pterygians, ichthyosauromorphs, thalatto-suchians, and mosasauroids. These analyseswere based on subsets of the primary data set(both continuous and binary characters). Somecharacters became redundant in the individualanalyses and were removed, and the datasubsets were then separately z-transformed


and ordinated using PCOa. Taxa wereassigned to new time bins covering the fullduration of each clade (SupplementaryTable 1). For each clade, disparity calculationswere based on the binned PCOa scores of taxafrom the ordination axes that accounted formore than 1% of absolute variance. Thenumber of axes used for each clade is asfollows: Sauropterygia (10), Eosauropterygia(11), Ichthyosauromorpha (7), Thalattosuchia(8), and Mosasauroidea (9). Once again, twodisparity metrics were examined: the sum ofvariances and rarefied sum of ranges.

We also calculated phylogenetic diversityestimates (PDE) for each group to explorethe pattern of numerical diversification(cladogenesis). PDE incorporate both taxonoccurrences and ghost lineages inferred fromtime-calibrated trees. Diversity counts weremade in 5Myr intervals spanning the durationof each of the five groups. To account forphylogenetic uncertainty within each tree,unresolved nodes were randomly resolvedand 100 trees were used to calculate PDE.The median of the 100 topologies was plottedalong with confidence intervals based on two-tailed 95% lower and upper quantiles. Calcu-lations were performed using the R packagepaleotree, Version 1.4 (Bapst 2012).Time-scaled Phylogenies.—Phylogenetic compa-

rative methods were used to quantify thetempo and mode of phenotypic evolution insauropterygians, eosauropterygians, ichthyosau-romorphs, thalattosuchians, and mosasauroids.Informal composite supertrees were firstconstructed for each clade (see descriptionin Supplementary Text and SupplementaryFigs. 2–5). All trees had unresolved nodes,reflecting phylogenetic uncertainty. Thesenodes were randomly resolved prior toanalyses, and to test for consistency, 50alternative fully resolved topologies wereretained for Ichthyosauromorpha, Thalatto-suchia, and Mosasauroidea, while 100topologies were analyzed for Sauropterygiaand Eosauropterygia. Branch durations in alltrees were estimated by assigning taxa apoint age, drawn randomly from a uniformdistribution between their first appearance dates(FAD) and last appearance dates (LAD). Absoluteages for FADs and LADs are from Gradstein

et al. (2012) (Supplementary Data). Zero-lengthbranches were lengthened by sharing durationequally with preceding non-zero-length branches,after setting a time of root divergence (equalmethod; Brusatte et al. 2008). Sensitivity analyseswere conducted with enforced 1Myr minimumbranch lengths (mbl method; Laurin 2004).Finally, prior to analyses, the taxa in eachfully resolved time-calibrated tree that lacked theappropriate trait data (see “Maximum-LikelihoodEvolutionary Models” below) were removed. Alldating procedures were implemented using thefunction timePaleoPhy in theRpackagepaleotree,Version 1.4 (Bapst 2012).

Maximum-Likelihood Evolutionary Models.—Maximum-likelihood models were fitted totrait data on the time-calibrated phylogeniesfor each marine reptile clade, using the Rpackage Geiger, Version 1.99-3 (Harmon et al.2008). PC scores from axes of variation wereused as continuous trait data to explore thetempo and mode of functional evolution in thejaws and dentition (e.g., Sallan and Friedman2012). The PC scores were taken from theseparate multivariate analyses of eachclade used to investigate disparity throughtime. Because no single ordination axisaccounts for all functional disparity, we fittedmodels to multiple axes independently. Theaxes utilized, and their relative contributionsto overall absolute variance, are: SauropterygiaPC1 (47.8%), PC2 (11.6%), and PC3 (5.7%);Eosauropterygia PC1 (30.5%), PC2 (13.7%),and PC3 (8.9%); Ichthyosauromorpha PC1(28.5%), PC2 (21.1%), and PC3 (13.2%);Thalattosuchia PC1 (44.6%), PC2 (18.1%), andPC3 (10.2%); and Mosasauroidea PC1 (30.6%),PC2 (24.8%), and PC3 (10.5%). Skull-sizetrends were also explored utilizing thelog10-transformed MSL data.

We focus on three models that explicitly testfor rate heterogeneity in a temporal context;Brownian motion (BM), early burst (EB), andPagel’s δ (delta). BM is a time-homogeneousprocess, equivalent to a random walk withconstant variance and evolutionary rates perunit time. Under BM, increases and decreasesin trait values are equiprobable, and traitcovariance is proportional to the duration ofshared ancestry. The EB model presumeshigher rates of evolution early in a clade’s


history and exponentially decreasing ratesthrough time (Harmon et al. 2010). In EB, ratesare constrained to exponential deceleration.BM and EB are the most commonly contrastedmodels when trying to identify the quantitativesignatures of niche filling during adaptiveradiation. Pagel’s δ is a time-dependent modelthat fits the relative contributions of earlyversus late evolution in a clade’s history byelaborating upon BM with an additionalphylogenetic rescaling parameter “δ” (Sallanand Friedman 2012). When Pagel’s δ< 1, traitdisparity and rapid morphological change areconcentrated early in a clade’s history,mimicking early burst (Ingram et al. 2012). Thesample size–corrected Akaike’s informationcriterion (AICc) and Akaike weights were usedto identify the best-fitting model. The averageand range of AICc values and Akaikeweights from each analysis, based on therandomly resolved time-calibrated topologies,are reported, in addition to the relevant para-meter values for favored models.

Results

Marine Reptile Morphospace.—Marine reptilefunctional morphospace visualizes thephenotypic similarity of the sampled taxabased on the functional data from thejaws and dentition (Fig. 2). Statistical testsmeasuring the strength of association betweenthe ordination axes and the functionalcharacters are summarized in SupplementaryTable 2. Here, discussion and visualization islimited to patterns on the first threemultivariate coordinate axes.

Together, PC1–PC3 account for 50% ofabsolute variance and encapsulate manysignificant traits (Fig. 2). PC1 subsumes 27.4%of absolute variance and 34.7% of variationwhen only axes with positive eigenvalues areconsidered. PC1 is most heavily influenced bydental specializations. Taxa with high nega-tive PC1 scores either possess toothplates (placodonts), are entirely edentulous(e.g., turtles), or have bulbous crushingdentition (e.g., mixosaurid ichthyosaurs andthalattosaurs). Conversely, taxa with lowpositive PC1 scores have pointed and recurvedtooth crowns, and those at the extreme positive

end of PC1 have specialized cutting dentition(Massare 1987). Most other characters alsohave strong association with positioning alongPC1, with the notable exception of jaw depth,area for muscle attachment, and anteriormechanical advantage, suggesting there ismarked variation in robustness and potentialbite force along this axis. PC2 accounts for12.1% of overall absolute variance and 15.3% ofvariation when only axes with positiveeigenvalues are considered. Variation inmandibular robustness and mechanicaladvantage is strongly expressed on PC2. Taxawith moderate to high negative PC2 scoreshave very slender jaws, with restricted regionsof attachment for the adductor muscles andlow anterior mechanical advantage. Taxa withincreasingly positive PC2 scores haveprogressively more robust jaws, with increasedarea for muscle attachment and higher anteriormechanical advantages. PC3 encapsulates afurther 10.7% of absolute variance and 13.5% ofvariation from positive eigenvalues only. Thisaxis appears to specifically isolate taxa withspecialized dentition in the anterior part of thedentary, such as enlarged anterior fangs orprocumbent chisel-like anterior teeth.

A number of noteworthy trends are evidentfrom the distribution of marine reptiles infunctional morphospace. Sauropterygiansoccupy an extensive area, but many formscluster centrally (Fig. 2). The sauropterygiansshown to be distinct from all other groupsin PC1–PC2 morphospace are the highlyspecialized placodonts. Most ichthyosaur-omorphs have restricted morphospaceoccupation along PC2, with slender jaws andlow mechanical advantages, but show widedistribution along PC1. Thalattosaurs are con-centrated in areas of weakly negative PC1scores, but a number of disparate forms branchout from this region. Turtles have restrictedmorphospace occupation, most likely due tolack of dental diversity (all edentulous, withthe exception of Odontochelys) and similargeometric configuration of the mandible—generally robust with a small or absentretroarticular process. In both plots, mostmosasauroids are concentrated together,potentially constrained by the presence of anintramandibular joint and unfusedmandibular


symphysis, together with moderately robustjaws and relatively high mechanical advan-tages. Thalattosuchians are positioned cen-trally in functional morphospace and showpartial overlap with other clades, particularlythe sauropterygians and ichthyosauromorphs.Tests for group separation in morphospace,based on NPMANOVA, return statisticallysignificant results for all pairwise comparisons(Supplementary Table 3).Mesozoic Marine Reptile Disparity.—Marine

reptile functional disparity was exceptionallyhigh in the Middle and early Late Triassic, butfollowing the Late Triassic extinctions, it tookover 100Myr to recover comparable levels of

disparity (Fig. 3A,B). This trend is present inboth disparity metrics. In the Olenekian,disparity is already high, and it increasesfurther into the Anisian (Middle Triassic).This demonstrates that, even in the context ofthe entire Mesozoic, marine reptiles achievedhigh levels of phenotypic variation within thefirst 10Myr of their evolution following thePTME. Greatest disparity is seen in the Carniantime bin in the early Late Triassic. With theexception of the Late Jurassic (Kimmeridgian–Tithonian) bin, disparity is consistently lowfrom the Early Jurassic to the Early Cretaceous.Disparity then increases from the Early toLate Cretaceous, and by the Campanian and

FIGURE 2. Empirical functional morphospaces showing the distribution of all marine reptile species and major groups.Two-dimensional plots of PCOa axes 1 and 2 and PCOa axes 2 and 3. The lower plots (B) represent the same axes butwith major groups denoted by convex hulls. The gray filled diamonds denote an unrelated assemblage of Triassicmarine reptiles. The illustrated jaws in PC1–PC2 plot are: Placodus (sauropterygian), Tylosaurus (mosasaur),Metriorhynchus (thalattosuchia), Ophthalmosaurus (ichthyosauromorph), Hupehsuchus (ichthyosauromorph), andNichollsemys (turtle). The illustrated jaws in PC2–PC3 plot are: Placodus (sauropterygian) and Nothosaurus(sauropterygian).


Maastrichtian, levels of variation equal (sum ofvariances) or approach (sum of ranges) those ofthe Middle Triassic and Carnian. Overall, thissuggests that marine reptiles were mostdisparate during the early and latter stages ofthe Mesozoic. Disparity and time-bin samplesize are very weakly and nonsignificantlycorrelated in the sum of variances (Spearman

rank correlation, ρ= 0.182, p= 0.515), butthere is a strong correlation between samplesize and disparity based on the sum ofranges when rarefied to the average samplesize (Spearman rank correlation, ρ= 0.807,p= 0.001) (Fig. 4).

There is evidence for a marked reductionin disparity resulting from marine reptile

FIGURE 3. Mesozoic marine reptile functional disparity. Mean disparity values based on the sum of variances (A) andsum of ranges (B) metrics (white circles) are plotted in 16 Mesozoic time intervals. The blue envelopes represent 95%confidence intervals based on 10,000 bootstrap replicates. The sum of ranges is rarefied to the average sample size ofthe 16 bins (n= 17). Partial disparity is illustrated in plot (C). This graphic illustrates the relative contributions of eachmarine reptile group to overall disparity through the Mesozoic: sauropterygians (green), ichthyosauromorphs (darkblue), thalattosaurs (orange), thalattosuchians (yellow), turtles (light blue), and mosasauroids (red).


extinctions throughout the Late Triassic. Whenexamining disparity in terms of the sum ofvariances metric (Fig. 3A), the overlappingconfidence intervals associated with the Norianand Hettangian–Sinemurian bins suggest anonsignificant disparity decline during theTriassic/Jurassic transition. This most likelyarises from inflated confidence intervals becauseof a low sample size in the Norian (n=5). Whenthis successive bin comparison is examinedusing variance-based statistics (Table 3),there is evidence for a statistically significant

decline, according to marginal likelihoods(LR: 8.27, exceeding threshold for significance;Royall 1997) and standard statistical tests(paired-sample t-tests, p=0.023; note this is notrobust to correction for multiple comparisons).The sum of ranges metric (Fig. 3B) showsa statistically significant decline between and theCarnian and Norian bins, but there is nodecline in disparity between the Norian andthe Hettangian–Sinemurian bin of the earliestJurassic, so conflicting with the results for thesum of variances. In this primary analysis, the

FIGURE 4. Mesozoic marine reptile disparity and time-bin sample size. Disparity through time is based on the data inFigure 3A,B, showing the mean sum of variances (solid black line) and sum of ranges (dashed black line) results. Binsample size is plotted in the same 16 Mesozoic time intervals (solid red line).

TABLE 3. Statistical tests for significant differences/shifts in functional disparity and functional morphospaceoccupation between successive Mesozoic time bins. For comparative purposes, statistical tests for a disparity shiftbetween the Carnian and Hettangian–Sinemurian are also provided. Disparity tests are based on paired-sample t-testsand likelihood ratios (LR). Functional morphospace occupation tests are based on nonparametric multivariate analysisof variance (NPMANOVA), performed on PC scores from the first 11 PC axes. Uncorrected and Bonferroni-correctedp-values are reported. Bold values represent statistically significant results where p-values are <0.05 and LRs are >8.See text in the “Stratigraphic Binning” section for full time-bin names and age ranges.

Paired-sample t-tests Marginal likelihoods NPMANOVA

Time-bin comparison p-value Corrected p-value Likelihood ratios (LR) p-value Corrected p-value

Ole/Ani 0.443 1 1.295 0.126 1Ani/Lad 0.520 1 1.102 0.982 1Lad/Car 0.232 1 1.861 0.168 1Car/Nor 0.631 1 1.003 0.304 1Nor/Het-Sin 0.023 0.373 8.268 0.001 0.108*Car/Het-Sin 0.011 0.177 10.374 — —Het-Sin/Plei-Toa 0.058 0.926 2.228 0.959 1Plei-Toa/Aal-Bath 0.489 1 1.025 0.493 1Aal-Bath/Cal-Oxf 0.396 1 1.125 0.569 1Cal-Oxf/Kim-Tith 0.100 1 12.751 0.242 1Kim-Tith/Ber-Bar 0.153 1 2.273 0.899 1Ber-Bar/Apt-Alb 0.072 1 13.592 0.337 1Apt-Alb/Cen-Tur 0.834 1 1.085 0.043 1Cen-Tur/Con-San 0.018 0.286 2.971 0.058 1Con-San/Camp 0.404 1 1.071 0.922 1Camp/Maas 0.361 1 1.197 0.278 1


sum of ranges is rarefied to the average samplesize of the 16 bins used (n=17). As the Norianbin has onlyfive samples, it is unsurprising that alarge reduction in disparity is recoveredwhen compared with the Carnian bin (rarefiedsample size is 17), because the sum of rangesmetrics is susceptible to sample-size bias. If allbins are rarefied to a minimum samplesize (n=5), there is no significant decline indisparity between the Carnian and Norian(Supplementary Fig. 6).

Marine reptile diversity was alreadymassively depleted in the Norian (Kelley et al.2014). Therefore, greater insights into the loss ofdisparity during the Late Triassic/Early Jurassictransition can be gained by comparing theCarnian bin with the Hettangian–Sinemurianbin. In this case, both variance- and range-basedmetrics show a statistically significant loss ofdisparity (Fig. 3A,B), and statistical tests confirmthat this disparity decline is significant, based onratios of marginal likelihoods for variance(LR: 10.37) and paired-sample t-tests (p=0.011)(Table 3). In conclusion, these analyses confirmthat the progressive and widespread lineageextinctions during the Late Triassic resultedin a significant loss of disparity in Mesozoicmarine reptiles.

Marine reptile disparity also declinedduring the Jurassic/Cretaceous transition. Bothmetrics show that disparity was higher in theLate Jurassic bin (Kimmeridgian–Tithonian)than at any other interval in the Jurassic, butdisparity was reduced by the Early Cretaceous(Berriasian–Barremian) (Fig. 3A,B). Marginaloverlap of the confidence intervals in the sum ofvariances metric suggests that the disparitydecline was nonsignificant. Variance-basedstatistical tests also fail to identify a significantchange in disparity between these two bins(LR: 2.27, paired-sample t-tests p=0.153).In contrast, confidence intervals associatedwith the sum of ranges metric do notoverlap. However, it is important to acknow-ledge sample-size discrepancies, because theBerriasian–Barremian bin has eight samples,compared with a rarefied sample size of 17 forthe preceding Late Jurassic bin. Overall, thereis a disparity reduction resulting from faunalturnover and a putative extinction duringthe Jurassic/Cretaceous transition, but the

magnitude and statistical significance of thisdecline is uncertain.

Partial Disparity Trends.—As a result ofextinctions, diversifications, and faunalturnover, the relative contribution of eachmarine reptile group to overall disparity trendsfluctuates markedly throughout the Mesozoic.Examining partial disparities dissects thecontributions of each of the six major taxonomicassemblages (Fig. 3C). In the earliest sampledinterval, the Olenekian, ichthyosauromorphsare by far the greatest contributors to overallfunctional disparity, but during the MiddleTriassic to early Late Triassic, sauropterygiansbecome the major contributors (approximately60–70%). Thalattosaurs remain consistentlylow contributors throughout the Triassic.Following the Late Triassic extinctions,ichthyosauromorphs (neoichthyosaurians) andsauropterygians (plesiosaurs) make equal contri-butions to disparity in the Early Jurassic.While the sauropterygian contribution remainsrelatively stable throughout the Middle and LateJurassic, the proportional contribution of theichthyosaurs becomes increasingly diminished.This is associated with a greater contribution todisparity by thalattosuchian crocodylomorphsand the diversification of plesiochelyid turtles inthe Late Jurassic. In the Early Cretaceous,ichthyosaurs are significant contributors todisparity, but their relative importance wanessubstantially before eventual extinction by theend-Cenomanian. Trends of partial disparityfrom the mid-Cretaceous onward are driven bythe diversification of the highly disparate turtlesand mosasauroids. The high levels of overallvariation in the Late Cretaceous coincide withthe dominance of these two clades. Theproportional disparity of the remaining LateCretaceous sauropterygians (polycotylids andelasmosaurs) is much reduced.

Temporal Morphospace Trends andSelectivity.—High disparity in the MiddleTriassic results from an early proliferation ofmorphospace occupation during the initialdiversifications of sauropterygians and ichthyo-sauromorphs, resulting from significantexcursions along the first two axes of variation(Fig. 5; see also Supplementary Fig. 7). Thisgross pattern of morphospace occupation ismaintained into the Carnian, and some


disparate functional morphotypes persist intothe Norian, despite reduced diversity. In theJurassic, morphospace occupation is relatively

restricted and consistent. As highlighted by thepartial disparity analyses (Fig. 3C), the clearincrease in disparity between the Early and Late

FIGURE 5. Patterns of functional morphospace occupation for marine reptiles through the Mesozoic. Two-dimensional plotsof PCOa axes 1 and 2 are illustrated for nine sampled intervals: Anisian, Carnian, Norian, Hettangian–Sinemurian, Aalenian–Bathonian, Kimmeridgian–Tithonian, Berriasian–Barremian, Cenomanian–Turonian, and Campanian. Symbols are used torepresent the major groups. The temporal position of each sampled interval is illustrated in a disparity through time plotbased on the sum of variances and sum of ranges (dotted line) metrics. All 16 intervals are figured in Supplementary Figure 7.


Cretaceous is overwhelmingly driven bydiversifications of the morphologically distinctmarine turtles and mosasauroids. The LateCretaceous is characterized by an abundance ofmosasauroids positioned at the positive extremeof PC1 and an absence of taxa with stronglynegative scores on PC2 (Fig. 5).

Statistical tests highlight two shifts in func-tional morphospace occupation between suc-cessive sampling intervals (Table 3). The firstsignificant shift is seen between the Norian andHettangian–Sinemurian bins, namely thoseassociated with the Triassic/Jurassic transition(NPMANOVA, F= 4.278, p= 0.0009). Thisresulted from the extinction of placodontsauropterygians and the excursion into regionsof morphospace marked by positive PC1 andPC2 scores by plesiosaurs in the Early Jurassic(Fig. 5). A second, less significant shift is foundbetween the Aptian–Albian and Cenomanian–Turonian bins of the mid-Cretaceous(NPMANOVA, F= 4.278, p= 0.0429, notrobust to p-value corrections for multiplecomparisons). This second shift could be dri-ven by the diversification of mosasauroids inthe Cenomanian and Turonian and thedecreasing abundance of ichthyosaurs.

There is evidence for ecologically selectiveextinctions through the Late Triassic. Whenmorphospace occupation in the Triassic bins iscompared with the Early Jurassic, it is evidentthat extinction victims are concentrated in theleft-hand regions of the plots (correspondingto negative PC1 scores) (Fig. 5). Throughoutthe Triassic, this adaptive zone was occupiedby various eosauropterygians, placodonts,thalattosaurs, and ichthyosaurs. These taxahad a range of functional specializations. Thosewith positive PC2 scores (i.e., placodonts andthalattosaurs) possessed specialized dentition,including crushing dentition and dentary toothplates, as well as having robust jaws withhigh mechanical advantages and increasedmusculature. Taxa with negative PC2 scoreshad slender jaws with weaker bites andreduced musculature, but all had dentalspecialization, including bulbous dentition(e.g., mixosaurids and Wumengosaurus)and in extreme cases, no dentition (e.g.,Endennasaurus). No marine reptiles in the EarlyJurassic possessed this suite of functional traits,

FIGURE 6. Marine reptile functional disparity plottedthrough time. Based on (A) within-bin mean pairwisedissimilarity calculated from the Gower intertaxondistance matrix using all characters, (B–E) using PC scoresfrom analyses using only continuous characters. Meandisparity values based on pairwise dissimilarity (A), andthe sum of variances (B, C) and sum of ranges (D, E)metrics (white circles) are plotted in 16 Mesozoic timeintervals. The blue envelopes represent 95% confidenceintervals based on 10,000 bootstrap replicates. The sum ofranges is rarefied to the average sample size of the 16 bins(n= 17). In (C) and (E) the character total mandibularlength was excluded from the data set.


and evolution is concentrated in a narroweradaptive zone (Fig. 5).Ancillary Disparity Results.—Disparity trends

are consistent when alternative protocols anddata subsets are used (Fig. 6). Based on meanpairwise dissimilarity from the Gowerintertaxon distance matrix, temporal patternsfollow those from the primary sum of variancesresults (Fig. 6A). Disparity trends based oncontinuous characters only also match those ofthe primary analyses. Once again disparitypeaks in the Middle and early Late Triassic.This trend is recovered in both metrics and withthe inclusion/exclusion of total mandibularlength (Fig. 6B–E). Disparity dynamics throughthe Triassic/Jurassic transition and Jurassic/Cretaceous transition follow those describedfrom the primary data set; the only exception isa lack of overlap between the confidenceintervals associated with the Jurassic/Cretaceous transition in the sum of variances(Fig. 6B,C). The only marked difference betweenthe primary results and ancillary results regardsthe extent to which disparity increased in theLate Cretaceous. Based on continuous charactersonly, levels of disparity reach a moderately highplateau from the Aptian–Albian bin until theend-Cretaceous; there is no apparent increase indisparity seen in the last three sampled intervalsof the Cretaceous. This result indicates thatdiscrete characters may be inflating disparity inthese latter bins. This is unsurprising, given thatthey are dominated by mosasaurs, whichpossess a unique intramandibular joint andhave great dental diversity.Patterns of Skull-Size Evolution.—The first

40Myr of marine reptile evolution witnessedan exceptional range of skull sizes. Collectively,the marine reptiles of the Triassic explored thefull range of forms seen in the entire Mesozoic(Fig. 7). Only a limited range of small forms ispresent in the Olenekian. However, theAnisian witnessed a great burst of skull-sizeevolution, including the diversification of bothsmaller taxa, such as pachypleurosaurs, andlarger forms, including gigantic nothosaursand the ichthyosaurs Cymbospondylus andThalattoarchon. This disparate array of cranialsizes is present just 5–10Myr after the PTME.Despite a considerable reduction in diversityduring the Norian, the overall range of skull

sizes remains high, including smaller forms suchas Endennasaurus and the largest marine reptileof the Mesozoic, the ichthyosaur Shonisaurussikanniensis (skull length ~3m).

The diversification of marine reptiles in theEarly Jurassic did not produce the same array offorms as in the post-PTME radiation (Fig. 7).Compared to the Triassic, the overall range ofsizes seen through the Jurassic is greatlyreduced. There is a lack of smaller forms but anabundance of large taxa, including ichthyosaurs,pliosaurids, and thalattosuchians. The range ofskull sizes is larger in the Late Jurassic, owing tothe diversification of plesiochelyid turtles andthe presence of large thalassophonean pliosaurs(e.g., Pliosaurus kevani, skull length ~2m).

A great diversity of skull sizes is found in themid-Cretaceous (Aptian–Turonian). Duringthis interval, the range of sizes equals thatof the Middle and Late Triassic (Fig. 7). Bothlarge-skulled pliosaurs (e.g., Kronosaurus,skull length ~2.3m) and small marine turtlesare found in the Aptian–Albian. Similarly,the Cenomanian–Turonian witnesses thediversification of marine squamates, some ofwhich (e.g., dolichosaurs) are very small,comparable to pachypleurosaurs of the Triassic(skull lengths of <10 cm). Large pliosaurs persistinto the Turonian (e.g., Megacephalosaurus, skull

FIGURE 7. Temporal trends of marine reptile skull-sizeevolution. In the upper plot, log10 skull length for 354marine reptile species is plotted at the midpoint of theirstratigraphic range. Symbols are used to differentiate themajor groups. Lower plot represents the same dataexpressed as box-and-whisker diagrams plotted at themidpoint of each time bin. Group symbols correspond toFigures 2 and 5.


length 1.8m). By the Late Cretaceous,mosasaurssuch as Tylosaurus, Prognathodon, Hainosaurus,and Mosasaurus show the largest sizes.

Clade-Specific Diversity and DisparityTrends.—Phylogenetic diversity estimatesshow variable trajectories for the five majormarine reptile clades. Sauropterygians,eosauropterygians, and ichthyosauromorphsachieve maximum species richness in theearliest intervals of their history (Fig. 8). Allthree clades show a significant decline inlineage diversity during the Late Triassic butgo on to achieve high diversity again in laterintervals, in the Early Jurassic (ichthyosaurs)and mid-Cretaceous (sauropterygians). For

mosasauroids, lineage diversity peaks in theCampanian, in the latter stages of theirevolutionary history, and there is no evidencefor an early numerical diversification. Thesame is true for thalattosuchians, which reachmaximum lineage diversity in the middle partof their duration (Callovian–Kimmeridgian)(Fig. 8).

Not all marine reptile clades show earlyhigh disparity. Calculations, based onseparate analyses of the five most diversemarine reptile clades, once again highlight adichotomy between those that diversifiedduring the Triassic biotic recovery and thosethat diversified later in the Mesozoic (Fig. 8).

FIGURE 8. Temporal diversity and disparity trends in five marine reptile groups. Phylogenetic diversity estimates areplotted in the first column. Mean disparity values (white circles) are plotted in time bins through each group’s duration,with associated 95% confidence intervals based on 10,000 bootstrap replicates. Two disparity metrics are shown:the sum of variances and the sum of ranges. The groups plotted are sauropterygians, eosauropterygians,ichthyosauromorphs, mosasauroids, and thalattosuchians. In the sum of ranges metric, the sample size is rarefied ton= 7, representing the median sample size for sauropterygians, eosauropterygians, ichthyosauromorphs, andthalattosuchians and the minimum sample size for mosasauroids. In column four, temporal trends of skull-sizediversity based on log10 skull length are plotted at the midpoint of each taxon’s stratigraphic range. Note thecontrasting temporal duration of each clade given on the x-axes.


Both sauropterygians and ichthyosaur-omorphs have bottom-heavy disparity pro-files, with greatest disparity present in theearly stages of their history and reduceddisparity in later intervals. In ichthyosaur-omorphs, this trend is still recovered when theenigmatic Cartorhynchus and Hupehsuchus areremoved (based on Ichthyosauria only). Forsauropterygians, this result is consistent whenanalyses are performed on eosauropterygiansonly (minus placodonts), but the magnitudeof difference between the early and laterintervals is far less (Fig. 8). Clades that diver-sified in the Jurassic and Cretaceous, thethalattosuchian crocodylomorphs and mosa-sauroids, have more top-heavy disparityprofiles, with reduced disparity in earlierintervals and greatest disparity later in theirdurations (Fig. 8).

Temporal trends of skull-size evolution alsoshow early bursts of evolution in sauropter-ygians and ichthyosauromorphs (Fig. 8).In both clades, an explosive diversificationgave rise to a great range of sizes in the earlystages of their evolutionary history (note thelogarithmic scale). Ichthyosauromorphs showa marked canalization of sizes through time,while sauropterygian evolution generallybecomes concentrated around exploring largerforms, spread over a greater time span. Incontrast to disparity trends, a relatively highrange of sizes is shown by thalattosuchiansduring their initial diversification; however,maximum range is not seen until the LateJurassic, ~30Myr after the clade’s origin.Closely corresponding with morphologicaldisparity trends, a limited range of generallysmaller forms is seen in mosasauroids for thefirst half of their history, before reaching astable higher range for the second half of theirevolution (Fig. 8).Clade-Specific Evolutionary Models.—Early-

burst maximum-likelihood models receiveoverwhelming support for sauropterygians,eosauropterygians, and ichthyosauromorphs,but not for mosasauroids and thalattosuchians(Fig. 9, Table 4). In the three clades thatdiversified during the Triassic biotic recovery,the early-burst model and/or delta model(with early high rates: δ< 1) have significantlygreater AICc weights for all phenotypic

variables (morphospace axes and skull size)(Fig. 9, Table 4). In contrast, BMmodels receivelittle support in these three clades. Inmosasauroids, models associated with rateheterogeneity have generally low AICcweights and are poorly supported, while theBMmodel is favored. For PC axis 2, the favoreddelta model is consistent with high rates ofevolution later in the clade’s history (δ> 1); theopposite of early burst (Ingram et al. 2012). TheEB model is best supported for skull-sizeevolution in mosasauroids, but this is notdecisive. In Thalattosuchia, EB model supportis mixed. Early burst and/or delta are clearlyfavored for skull-size evolution, but a BMmodel is best supported for functionalevolution in the jaws and dentition (Fig. 9,Table 4). When dating trees with 1Myrminimum branch lengths enforced, the resultsare the same (Supplementary Table 4).

Discussion

Early Ecomorphological Diversification in TriassicMarine Reptiles.—As an adaptive assemblage,Triassic marine reptiles were one of the truesuccess stories to arise in the devastatedpostextinction oceans. The first wave ofmarine invasion by reptiles in the Triassicwas associated with an exceptional burst ofecomorphological diversity (Figs. 3, 5, 7).Taken together, patterns of functionaldisparity, morphospace occupation and skull-size evolution, show that the Middle to earlyLate Triassic was not just a time of markedproliferation in terms of species numbers (e.g.,Benson et al. 2010; Benson and Butler 2011), butalso a time of explosive phenotypic evolution—something that had only previously been notedspeculatively based on qualitative observations(Fröbisch et al. 2013; Liu et al. 2014). In the first10–20Myr of the Triassic, diversifying lineagesexplored the greatest breadth of functionalmorphospace (Fig. 5), and the accumulatedfunctional disparity represented the maximumin the Mesozoic (Fig. 3). Similarly, the Triassicwitnessed the greatest disparity of skull sizes,with the full range of potential forms realized byone clade or another (Fig. 7).


The geologically rapid evolution of suchphenotypic diversity likely resulted fromearly expansion into a broad variety ofpreviously unexplored trophic niches thatrepresented vacant adaptive zones (Bentonet al. 2013). Diverse diets, including feedingon fast-moving nektonic prey, hard- andsoft-shelled benthic invertebrates, and othermarine reptiles, resulted in many specializedmorphofunctional innovations, including robustjaws with massive coronoid processes, powerfulbites and crushing dentition, benthic feederswith specialized anterior dentition to acquireprey from the substrate, snapping bitesgenerated with gracile jaws accommodatingprocumbent fangs, and large macropredatorswith enlarged bicarinate cutting teeth. Fornew top-tier predatory tetrapods, a marinerevolution was already well underway in theMiddle Triassic. Our study points to an earlier

onset of the Mesozoic marine revolution (MMR)in some groups compared with the usual begin-ning point of the Jurassic (Vermeij 1977). Thisearly onset of the MMR is seen among marinereptiles and is also suggested by new discoveriesof lobsters and neopterygian bony fishes fromthe Anisian in China (Hu et al. 2011) and evi-dence of a predation-driven Triassic radiation incrinoids (Baumiller et al. 2010). Overall, this hintsthat theMMRmay have been triggered as part ofthe recovery from the PTME, rather than later inthe Jurassic (Vermeij 1977).

Much recent research effort has focused onthe tempo of biotic recovery following thePTME. Was the full global recovery of marinelife delayed until 5–10Myr after the event, bywhich time complex and stable ecosystemswere established (Chen and Benton 2012; Liuet al. 2014), or was the recovery of life in theoceans more rapid, with full-length, multilevel

FIGURE 9. Evolutionary model fittings for morphospace axes and skull size in five marine reptile groups. Akaike weights ofthree models are expressed as circle charts for each group and trait. The groups plotted are: sauropterygians,eosauropterygians, ichthyosauromorphs, mosasauroids, and thalattosuchians. In all instances where delta is the favored model,it is associated with early high rates, with the exception of mosasauroid PC axis 2. EB, early burst; BM, Brownian motion.


TABLE 4. Summary of maximum-likelihood model-fitting analyses. Sample size–corrected Akaike’s information criterion (AICc), model parameter values, and AICc weightsare documented for each analytical permutation. For the AICc values, we report the mean and standard error based on multiple dating and topology replicates, as describedin the text. The results from this table are visualized in Figure 9. The best-fitting evolutionary models are highlighted in bold for each analysis: lowest AICc value, highestAICc weight. MSL, maximum skull length; BM, Brownian motion; and EB, early burst.

Clade and data Variables EB Delta BM

Sauropterygia PC1 AICc −179.62SE: 2.17

−184.17SE: 3.11

−122.81SE: 0.93

AICc weight 0.093 0.907 0.000Parameters −0.027 0.175 —


−176.41SE: 2.41

−149.48SE: 1.26



−203.88SE: 3.35

−182.39SE: 2.14


Sauropterygia log10 MSL AICc 100.21SE: 6.06

98.36SE: 7.680

173.83SE: 6.41


Eosauropterygia PC1 AICc −92.32SE: 1.09

−88.14SE: 1.79

−48.48SE: 0.89



−126.22SE: 2.55

−125.02SE: 2.49



−100.47SE: 3.15

−91.38SE: 2.39


Eosauropterygia log10 MSL AICc 85.77SE: 6.26

85.65SE: 7.88

163.46SE: 6.11


Ichthyosauromorpha PC1 AICc −59.01SE: 1.03

−55.94SE: 1.10

−25.62SE: 1.50



−51.26SE: 2.00

−20.45SE: 2.12


MESO

ZOIC

MARIN

EREPT

ILEDISPA

RITY

19

Table 4. Continued

Clade and data Variables EB Delta BM


−73.45SE: 2.00

−39.95SE: 2.35


Ichthyosauromorpha log10 MSL AICc 19.23SE: 1.74

25.19SE: 1.93

50.56SE: 2.16


Mosasauroidea PC1 AICc −40.89SE: 0.38

−40.43SE: 0.35

−42.71SE: 0.4

AICc weight 0.234 0.186 0.581Parameters — — —


−43.96SE: 0.36

−41.32SE: 0.47

AICc weight 0.057 0.744 0.199Parameters — 2.263 —


−67.83SE: 0.37

−69.91SE: 0.34


Mosasauroidea log10 MSL AICc −18.96SE: 1.86

−17.86SE: 1.76

−18.42SE: 1.65


Thalattosuchia PC1 AICc −18.33SE: 0.36

−17.86SE: 0.38

−19.86SE: 0.45



−30.74SE: 0.91

−28.82SE: 1.05

AICc weight 0.864 0.0982 0.0376Parameters −0.087 — —


−39.11SE: 0.32

−41.65SE: 0.35


Thalattosuchia log10 MSL AICc −42.48SE: 0.93

−42.18SE: 1.04

−38.19SE: 1.22


20THOMASL.ST

UBBSAND

MIC

HAELJ.B

ENTON

trophic food webs present in the Induan orOlenekian (Scheyer et al. 2014; Motani et al.2015a)? Although this study was not designedto test these competing hypotheses, the resultscan still be used to inform this debate.

Analyses show that marine reptilefunctional disparity was already relatively highin the Olenekian, particularly in terms of var-iance, when ichthyosauromorphswere dominantcontributors to overall disparity (Fig. 3). Thisresult is intriguing, given low species diversity(Kelley et al. 2014), but it is not entirelyunexpected when surveying the global diversityof forms. Olenekian marine reptiles had alreadybroadly diversified into an array of feedingstrategies, including suction-feeding andlunge-feeding ichthyosauromorphs such asCartorhynchus and Hupehsuchus, potentiallydurophagous ichthyosaurs like Grippia, andearly presumed fish-eating eosauropterygianslike Corosaurus (Motani 1997; Rieppel 1998;Motani et al. 2015a,b). Disparity increased intothe Anisian, when sauropterygians becamedominant contributors to overall disparity(Figs. 3, 5), although the magnitude of increasewas not statistically significant. A moresubstantial increase in the diversity of formsbetween the Olenekian and Anisian is seen in theskull-size data. The Carnian bin represents thedisparity maximum, marginally greater than inthe Anisian and latest Cretaceous bins (Fig. 3).No new higher clades diversified during theCarnian, nor was the disparity peak associatedwith a considerable expansion of morphospaceoccupation or shift in partial disparitycontributions (Figs. 3, 5). It is likely that thehigh Carnian disparity resulted from the accu-mulated sample of functional extremes, such aslarge edentulous ichthyosaurs, shell-crushingplacodonts (including the enigmatic Henodus),and bizarre heterodont thalattosaurs (e.g.,Xinpusaurus). Overall, these patterns cannotexclude either hypothesis regarding the tempo ofbiotic recovery following the PTME, but there istentative support for a delay until the Middle–early Late Triassic, when marine reptiles reachedtheir full potential of ecomorphological diversity.Late Burst in Cretaceous Marine Reptiles.—This

study highlights the Late Cretaceous as asecond exceptional interval in the trophicdiversifications of Mesozoic marine reptiles.

Functional disparity in the jaws and dentitionof Campanian and Maastrichtian taxa equaledthat of the Middle and early Late Triassic(Fig. 3). This resulted from the diversification ofdisparate mosasauroids and turtles, coupledwith the persistence of elasmosaurid andpolycotylid plesiosaurs. Mosasaurs becameincreasingly ecologically disparate through theirgeological history, culminating in broadecospace occupation in the Campanian andMaastrichtian (Ross 2009). The substantialdepth distribution of mosasaurs in the watercolumnmeant they could exploit diverse benthicand pelagic prey (Polcyn et al. 2014). Turtles alsoachieved great taxonomic diversity in the LateCretaceous and possessed functionally distinctrobust jaws without teeth (Brinkman et al. 2006).Massare (1987) suggested that there was areduction in marine reptile ecomorphologicaldiversity between the Late Jurassic and LateCretaceous, by examining dental morphotypesin the Kimmeridge Clay, Pierre Shale, andNiobrara Chalk faunas. However, this trendwas not recovered here, most likely because thecurrent study incorporates a greater diversity oftaxa from global samples (including turtles) andalso considers structural variation in the jaw.

There is no substantive evidence for alarge-scale decline in marine reptileecomorphological diversity prior to the end-Cretaceous. Therefore, any destructionbrought about by the end-Cretaceous massextinction was likely to have been geologicallyabrupt. To investigate further, future studiescould focus on narrower sampling intervalsand regional trends in the last 10Myr of theMesozoic.

Disparity, Lagerstätte Effects, and Shallow-Marine Environments.—Exceptionally richfossil lagerstätten can be overwhelmingdeterminants in temporal patterns oftaxonomic diversity, because atypically highpreservation can generate artificial peaks(e.g., Butler et al. 2013). In Mesozoic marinereptiles, diversity in the Jurassic, and to someextent the Middle Triassic, is heavily affectedby lagerstätte-dominated sampling (Bensonet al. 2010; Benson and Butler 2011). Theinfluences of “lagerstätte effects” on patternsof disparity are not well understood.Theoretically, formations with exceptional


preservation are more likely to providematerial of sufficient quality, such ascomplete skulls and skeletons, that can beincluded in large-scale studies of disparity,potentially inflating sample size in suchintervals. In fact, exceptionally preservedmarine reptile biotas that yield manyspecimens are scattered through theMesozoic, and it is a moot point which wouldbe termed “lagerstätten,” and which not.

A detailed study of the ichthyosaur fossilrecord showed that specimen quality is subjectto many factors (Cleary et al. 2015). Theseinclude geographic location (the NorthernHemisphere is better documented than theSouthern), specimen size (medium-sizedspecimens are more complete than small orlarge), and facies (best specimens infine-grained siliciclastics). Importantly, Clearyet al. (2015) found no relationship betweenspecimen quality and any of the commonlyused temporal sampling metrics such asformation counts or map areas, nor werenamed lagerstätten the unique sources ofcomplete specimens.

Disparity is widely considered to be bothconceptually and empirically different fromdiversity; sampled intervals commonlyhave high diversity but low morphologicaldisparity, or vice versa (Foote 1997) (Fig. 4).In addition, variance-based measures aregenerally robust to sample-size discrepancies(Ciampaglio et al. 2001). In this study, tem-poral patterns of functional disparity andskull-size variation cannot be simply attributedto the distribution of lagerstätte deposits.Jurassic marine reptiles together exhibitgenerally low disparity, despite beingdominated by lagerstätten (Figs. 3–7). Incontrast, the mid- to Late Cretaceous intervalshows higher disparity, despite not beingassociated with lagerstätte effects (Benson andButler 2011).

Marine reptile diversity trends were drivenby marine transgression and regression(Benson and Butler 2011). Marine reptiles canbe broadly divided into shallow-marine oropen-ocean habitat groups based on the degreeof postcranial specialization and locomotorymodes. Benson and Butler (2011) identified theAnisian–Carnian, Bathonian–Tithonian, and

Cenomanian–Maastrichtian as times withelevated diversity of shallow-marine taxa.These intervals broadly correspond to sea-levelhighstands and times of greater continentalflooding. The strong negative correlationrecovered between shallow-marine taxicdiversity and nonmarine area was interpretedby Benson and Butler (2011) as representing aspecies diversity–area relationship, wherebygreater continental flooding increases thehabitable area for shallow-marine organismsand elevates the deposition of fossiliferousrock, a pattern formalized as the “commoncause” hypothesis (Peters 2005). Intriguingly,times of transgression, increased continentalflooding, and higher diversity in shallow-marine taxa also correspond to intervals ofgreatest disparity in the current study.Functional disparity and the diversity ofskull sizes are highest in the Anisian–Carnian,Kimmeridgian–Tithonian and Late Cretaceous(Figs. 3, 7). Intervals in which shallow-marinereptiles are rarer and open-ocean taxadominated, such as the Early Jurassic andearliest Cretaceous (Benson and Butler 2011),have reduced functional disparity and a lessdiverse range of skull sizes. Such times occurduring or after major regression events (Bensonand Butler 2011; Kelley et al. 2014). This studytherefore provides tentative evidence that majorpatterns ofmorphological evolution inMesozoicmarine reptiles were driven by changingphysical environmental conditions.

Shallow-marine reptiles may exhibit greaterfunctional disparity and a more diverse rangeof skull sizes because coastal and shallow-shelf environments accommodate a greaterdiversity of habitats and of prey. Transgressionand continental flooding concentrate nutrientsin coastal and shallow-shelf environmentsbecause there is abundant terrigenous inputthrough sediments and soils (Smith et al. 2001).This is expected to increase productivity andbiomass, particularly in benthic invertebrates(Polycn et al. 2014). A diversity of prey is likelyto catalyze phenotypic innovation in thejaws and dentition. For example, feeding onhard-shelled benthic invertebrates requiresmorphofunctional modifications, such asincreased musculature, greater jaw robustness,higher mechanical advantages, crushing or


bulbous dentition, and specialized anteriordentition for plucking sessile prey.Durophagous and benthic foraging specializa-tions are abundant in the shallow-marinereptiles, including Triassic sauropterygians,thalattosaurs and ichthyosauromorphs, LateJurassic plesiochelyid turtles, and specializedCretaceous mosasauroids and turtles.

Deep-water facies, inhabited by open-oceanmarine reptiles, receive little terrigenous inputand are expected to be more resource poor(Hudson et al. 1991; Benson and Butler 2011).Open-ocean marine reptiles would generallyfeed on nektonic prey, such as fast-movingcephalopods, fish, and tetrapods (Massare1987), and therefore the absence of benthicforagers and durophages could explainreduced disparity. Hydrodynamic constraintson structural variation are also heightened inopen oceans when feeding on fast-movingnektonic prey occurs (Taylor 1987).Extinction Intensity and Selectivity.—Selective

extinction and the lack of opportunisticecospace refilling in the Jurassic meant thatthe Late Triassic extinctions had profoundconsequences for marine reptile evolution.Major marine regression through the lateCarnian and Norian, followed by massiveeruptions and ocean anoxia close to theTriassic/Jurassic boundary, resulted inwidespread lineage extinction, and theassociated decline in functional disparityand skull-size diversity identified here hadlong-lasting effects, with comparable levels ofecomorphological variation not recovered foranother 100Myr. Low sample size in thelong Norian bin and lack of material for theRhaetian makes assessing the timing ofthe extinctions and loss of disparity difficult.However, the functionally distinct placodontspersisted into the Rhaetian (Rieppel 2002;Nordén et al. 2015), so disparity could haveremained high until the Triassic/Jurassicboundary. Extinction victims were concentratedin the moderate to extreme negative regions ofPC1 in functional morphospace, an areaassociated with durophagous and benthicforaging specializations. This supports thefindings of Kelley et al. (2014), who, usingcategories, illustrated that the shallow-marinedurophagous taxa were vulnerable to the

geologically rapid regression of the LateTriassic. Our study also points toward size asan ecologically selective trait. Whereas moststudies find selectivity against larger taxa(e.g., Friedman 2009), in this case it was smallmarine reptiles that did not transcend theextinction interval (Fig. 7).

Major extinction events are predicted togive rise to episodes of morphologicaldiversification when ecological space is rapidlyrefilled during the recovery interval (Droseret al. 1997; Erwin 2008). Jurassic marine reptilesdo not conform to this trend but instead appearto have passed through a macroevolutionarybottleneck (Thorne et al. 2011). Althoughmarine reptile groups diversifying in the EarlyJurassic achieved high numerical diversity(Thorne et al. 2011; Benson et al. 2012; Clearyet al. 2015) (Fig. 8), they failed to explore thesame range of ecomorphological character-istics as extinction victims and were generallyconcentrated in a reduced range of adaptivezones (Figs. 3, 5, 7). Similar canalization wasreported by Thorne et al. (2011) and Dick andMaxwell (2015) for ichthyosaurs based onmorphological variation in whole-bodyskeletal characters and ecospace modeling.As previously noted, this could be attributed tovariable diversification patterns in shallow-marine versus open-ocean environments. Thediversifications of chondrichthyans andosteichthyans may have also impacted marinereptile macroevolution. For example, there areelevated origination rates in chondrichthyansduring the Early Jurassic (Friedman and Sallan2012), while others have highlighted greaterpotential competition in the small-bodied,durophagous, and benthic-foraging nichesfrom actinopterygians such as Dapedium(Thorne et al. 2011; Smithwick 2015).

There is some evidence for a loss ofecomorphological diversity through theJurassic/Cretaceous transition. Marine reptilefunctional disparity decreased, but statisticalsupport for this decline is generally weak orabsent (Fig. 3, Table 3). Representatives of allmajor groups passed through the extinctioninterval (Benson and Druckenmiller 2014),along with most functional morphotypes.

Adaptive Radiations in the Marine Realm.—When separately diversifying clades are


presented with the same ecologicalopportunity, such as invading oceanecosystems, evolution is expected to bestrongly deterministic, leading to replicateadaptive radiations and repeated bursts ofevolution (Schluter 2000). However, thisstudy shows that invasion of a new habitatdoes not lead to identical macro-evolutionary trajectories in the most diverseMesozoic marine reptile groups. Thesignatures of adaptive radiation are seen insauropterygians, eosauropterygians, andichthyosauromorphs, as shown universally bythe early accumulation of lineage diversity,early high disparity, rapid proliferation ofskull-size diversity, and strong supportfor early-burst maximum-likelihood models(Figs. 8, 9, Table 4). In contrast, an adaptiveradiation model does not adequately describepatterns of evolution in thalattosuchiansand mosasauroids. This suggests that thecolonization of a new environment alone doesnot always serve as a catalyst for adaptiveradiation and spectacular speciation events.

Timing of diversification and bioticconditions in the marine realm were likely keyfactors in the contrasting patterns of evolutionin Mesozoic marine reptiles. As stated bySimpson (1944: p. 212) “The availability of anew adaptive zone does not depend alone onits physical existence … , but also on its beingopen to other occupants or so sparsely ormarginally occupied that it involves no greatcompetition.” When sauropterygians and ich-thyosauromorphs diversified in the aftermathof the PTME, the marine realm was largelydevoid of competitors, leading to numericaland morphological diversifications alongdisparate ecological axes (Fig. 5) (Chen andBenton 2012). In contrast, thalattosuchians andmosasauroids originated free from anymajor biotic crises. Thalattosuchians initiallydiversified in the Early Jurassic (Young et al.2010), ~20Myr after secondary Jurassicradiations gave rise to a diverse assemblageof plesiosaurian sauropterygians andneoichthyosaurs (Thorne et al. 2011; Bensonet al. 2012). Mosasauroids diversified in theCenomanian and Turonian, when plesiosaurswere represented by three ecologically distinctgroups—the elasmosaurids, polycotylids, and

pliosaurids—and large predatory teleosts andsharks were important components of marineecosystems (Massare 1987; Schwimmeret al. 1997; Everhart et al. 2010). Therefore,for thalattosuchians and mosasauroids,diversifying selection during the initial phaseof evolution may not have been as strong,because there was less unoccupied niche spaceto fill. These divergent trends agree with thegrowing consensus on how ecologicalopportunity may operate to drive adaptiveradiation, through both “niche availability”(e.g., marine-environment prey resource) andthen “niche discordance,” which promotesphenotypic diversification into increased var-iance of niche-related traits (diverse diets andfeeding ecologies) (Schluter 2001; Burbrinket al. 2012; Wellborn and Langerhans 2015).

Clade duration and large-scale externalperturbations may also explain the dichot-omous macroevolutionary trends in Mesozoicmarine reptiles. Sauropterygians and ichthyo-sauromorphs differ from mosasauroids andthalattosuchians in terms of evolutionaryduration and terrestrial phylogenetic heritage.Thalattosuchians and mosasauroids representshorter-lived and specialized offshoots fromlarge, ancestrally terrestrial, clades (Bardet et al.2014). On the other hand, ichthyosauromorphsand sauropterygians were long-lived marineclades, and their evolution was punctuated bymajor phylogenetic bottlenecks in the LateTriassic (Thorne et al. 2011). Mosasaur extinc-tion was particularly abrupt and seeminglypremature, given that they maintained highdiversity and disparity in the later parts oftheir history (Polcyn et al. 2014) (Fig. 8). Hadmosasaurs passed through the Cretaceous–Paleogene extinction event and continued toexist with reduced diversity and disparityfor a long interval, their macroevolutionarytrends may have closely mirrored those ofsauropterygians and ichthyosauromorphs.

Direct tests for bursts of evolution character-istic of adaptive radiation are relatively rare forother secondarily marine tetrapod groups. TheCenozoic witnessed the marine diversificationsof sphenisciforms, hydrophiin snakes, cetaceans,sirenians, and pinnipedimorphs (Kelley andPyenson 2015). These independent invasions ofocean ecosystems were broadly analogous to


those seen in Mesozoic marine reptiles; mostwere initiated by ecological opportunity andoccurred in comparable environmental settings(Pyenson et al. 2014). Thus far,most attention hasfocused on the diversification of cetaceans. Openniche space has been proposed as the maindriving force behind the Eocene archaeocete(stem cetacean) radiation (Gingerich 2003).Steeman et al. (2009) and Slater et al. (2010) bothfocused on the tempo of the neocete (crowncetacean) radiation in a quantitative frameworkand discovered no evidence for early rapidlineage diversification in the Oligocene.Importantly, however, Slater et al. (2010) didpresent strong evidence for an early burst ofbody-size evolution in neocetes associated withdietary differentiation, consistent with a niche-filling adaptive radiation model. Therefore, thereis some evidence that empty ecospace was notjust a driver for the adaptive radiations of Trias-sic ichthyosaurs and sauropterygians but alsoimportant in the ecological ascent of cetaceans.Future studies may benefit from incorporatingboth stem and crown cetaceans with fossil datawithin the same framework to facilitate abroader comparison of trends across secondarilymarine tetrapods through their entire durations.

Acknowledgments

This project was funded by the NaturalEnvironment Research Council (Ph.D. grantNE/J500033/1 to T.L.S. and M.J.B.). We thankHeinz Furrer, Rainer Schoch, andMarkusMoserfor providing access to specimens and assistanceduring collection visits. We are very grateful toSven Sachs, Benjamin Moon, Judyth Sassoon,Benjamin Kear, Silvio Renesto, James Neenan,Davide Foffa, Mike Polcyn, Da-yong Jiang,William Simpson, Jahn Hornung, Carl Mehling,Anne Schulp, and Jan Ove R. Ebbestad forproviding photographs of specimens. Manythanks also to Marcello Ruta, Phil Anderson,and Emily Rayfield for methodological adviceand to two reviewers for helpful suggestionsand comments. Finally, we thank the Paleobiol-ogy Patrons Fund for financial support.

Literature CitedAnderson, P. S. L. 2009. Biomechanics, functional patterns, anddisparity in Late Devonian arthrodires. Paleobiology 35:321–342.

Anderson, P. S. L., M. Friedman, M. D. Brazeau, and E. J. Rayfield.2011. Initial radiation of jaws demonstrated stability despitefaunal and environmental change. Nature 476:206–209.

Anderson, P. S. L., M. Friedman, and M. Ruta. 2013. Late to thetable: diversification of tetrapod mandibular biomechanicslagged behind the evolution of terrestriality. Integrative andComparative Biology 53:197–208.

Bapst, D. W. 2012. Paleotree: an R package for paleontological andphylogenetic analyses of evolution. Methods in Ecologyand Evolution 3:803–807.

Bapst, D. W., P. C. Bullock, M. J. Melchin, H. D. Sheets, andC. E. Mitchell. 2012. Graptoloid diversity and disparity becamedecoupled during the Ordovician mass extinction. Proceedingsof the National Academy of Sciences USA 109:3428–3433.

Bardet, N. 1994. Extinction events amongMesozoic marine reptiles.Historical Biology 7:313–324.

Bardet, N., J. Falconnet, V. Fischer, A. Houssaye, S. Jouve,X. P. Suberbiola, and P. Vincent. 2014. Mesozoic marine reptilepalaeobiogeography in response to drifting plates. GondwanaResearch 26:869–887.

Baumiller, T. K., M. A. Salamon, P. Gorzelak, R.Mooi, C. G. Messing,and F. J. Gahn. 2010. Post–Paleozoic crinoid radiation in response tobenthic predation preceded the Mesozoic marine revolution. Pro-ceedings of the National Academy of Sciences USA 107:5893–5896.

Benson, R. B. J., and R. J. Butler. 2011. Uncovering the diversifica-tion history of marine tetrapods: ecology influences the effect ofgeological sampling biases. Geological Society of London SpecialPublication 358:191–207.

Benson, R. B. J., and P. S. Druckenmiller. 2014. Faunal turnover ofmarine tetrapods during the Jurassic–Cretaceous transition.Biological Reviews 89:1–23.

Benson, R. B. J., R. J. Butler, J. Lindgren, and A. S. Smith. 2010.Mesozoic marine tetrapod diversity: mass extinctions and temporalheterogeneity in geological megabiases affecting vertebrates.Proceedings of the Royal Society of London B 277:829–834.

Benson, R. B. J., M. Evans, and P. S. Druckenmiller. 2012. Highdiversity, low disparity and small body size in plesiosaurs(Reptilia, Sauropterygia) from the Triassic–Jurassic boundary.PLoS ONE 7:e31838.

Benson, R. B. J., N. E. Campione, M. T. Carrano, P. D. Mannion,C. Sullivan, P. Upchurch, and D. C. Evans. 2014. Rates ofdinosaur body mass evolution indicate 170 million years of sus-tained ecological innovation on the avian stem lineage. PLoSBiology 12:e1001853.

Benton, M. J., Q. Zhang, S. Hu, Z-Q. Chen, W.Wen, J. Liu, J. Huang,C. Zhou, T. Xie, J. Tong, and B. Choo. 2013. Exceptionalvertebrate biotas from the Triassic of China, and the expansion ofmarine ecosystems after the Permo–Triassic mass extinction.Earth-Science Reviews 137:85–128.

Benton, M. J., J. Forth, andM. C. Langer. 2014. Models for the rise ofthe dinosaurs. Current Biology 24:R87–R95.

Brinkman, D., M. Hart, H. Jamniczky, and M. Colbert. 2006.Nichollsemys baieri gen. et sp. nov, a primitive chelonioidturtle from the late Campanian of North America. Paludicola5:111–124.

Brusatte, S. L., M. J. Benton, M. Ruta, and G. T. Lloyd. 2008.Superiority, competition, and opportunism in the evolutionaryradiation of dinosaurs. Science 321:1485–1488.

Burbrink, F. T., X. Chen, E. A. Myers, M. C. Brandley, andR. A. Pyron. 2012. Evidence for determinism in speciesdiversification and contingency in phenotypic evolution duringadaptive radiation. Proceedings of the Royal Society of London B279:4817–4826.

Butler, R. J., R. B. J. Benson, and P. M. Barrett. 2013. Pterosaurdiversity: untangling the influence of sampling biases,Lagerstätten, and genuine biodiversity signals. Palaeogeo-graphy, Palaeoclimatology, Palaeoecology 372:78–87.


Button, D. J., E. J. Rayfield, and P. M. Barrett. 2014. Cranialbiomechanics underpins high sauropod diversity in resource-poor environments. Proceedings of the Royal Society of LondonB 281:20142114.

Chen, Z. Q., andM. J. Benton. 2012. The timing and pattern of bioticrecovery following the end-Permian mass extinction. NatureGeoscience 5:375–383.

Cheng, L., X. H. Chen, Q. H. Shang, and X. C. Wu. 2014. A newmarine reptile from the Triassic of China, with a highly specia-lized feeding adaptation. Naturwissenschaften 101:251–259.

Ciampaglio, C. N., M. Kemp, and D. W. McShea. 2001. Detectingchanges in morphospace occupation patterns in the fossil record:characterization and analysis of measures of disparity. Paleo-biology 27:695–715.

Cleary, T. J., B. C. Moon, A. M. Dunhill, and M. J. Benton. 2015. Thefossil record of ichthyosaurs, completeness metrics andsampling biases. Palaeontology 58:521–536.

Close, R. A., M. Friedman, G. T. Lloyd, and R. B. J. Benson. 2015.Evidence for a mid-Jurassic adaptive radiation in mammals.Current Biology 25:2137–2142.

Dick, D. G., and E. E. Maxwell. 2015. The evolution and extinctionof the ichthyosaurs from the perspective of quantitative ecospacemodelling. Biology Letters 11:20150339.

Droser, M. L., D. J. Bottjer, and P. M. Sheehan. 1997. Evaluating theecological architecture of major events in the Phanerozoic historyof marine invertebrate life. Geology 25:167–170.

Erwin, D. H. 2007. Disparity: morphological pattern anddevelopmental context. Palaeontology 50:57–73.

——. 2008. Extinction as the loss of evolutionary history. Proceed-ings of the National Academy of Sciences USA 105:11520–11527.

Everhart, M. J., S. A. Hageman, and B. L. Hoffman. 2010. AnotherSternberg “fish-within-a-fish” discovery: first report of Ichthyo-dectes ctenodon (Teleostei; Ichthyodectiformes) with stomachcontents. Transactions of the Kansas Academy of Science113:197–205.

Finarelli, J. A., and J. J. Flynn. 2007. The evolution of encephaliza-tion in caniform carnivorans. Evolution 61:1758–1772.

Foote, M. 1993. Contributions of individual taxa to overallmorphological disparity. Paleobiology 19:403–419.

——. 1997. The evolution of morphological diversity. AnnualReview of Ecology and Systematics 28:129–152.

Friedman, M. 2009. Ecomorphological selectivity among marineteleost fishes during the end-Cretaceous extinction. Proceedingsof the National Academy of Sciences USA 106:5218–5223.

——. 2012. Parallel evolutionary trajectories underlie the origin ofgiant suspension-feeding whales and bony fishes. Proceedings ofthe Royal Society of London B 279:944–951.

Friedman, M., and L. C. Sallan. 2012. Five hundred million years ofextinction and recovery: a Phanerozoic survey of large‐scalediversity patterns in fishes. Palaeontology 55:707–742.

Fröbisch, N. B., J. Fröbisch, P. M. Sander, L. Schmitz, and O. Rieppel.2013.Macropredatory ichthyosaur from theMiddle Triassic and theorigin of modern trophic networks. Proceedings of the NationalAcademy of Sciences USA 110:1393–1397.

Gingerich, P. D. 2003. Land-to-sea transition of early whales:evolution of Eocene Archaeoceti (Cetacea) in relation to skeletalproportions and locomotion of living semiaquatic mammals.Paleobiology 29:429–454.

Gower, J. C. 1971. A general coefficient of similarity and some of itsproperties. Biometrics 27:857–871.

Gradstein, F. M., J. G. Ogg, M. Schmitz, and G. Ogg. 2012. Thegeologic time scale 2012. Elsevier, Amsterdam.

Harmon, L. J., J. T. Weir, C. D. Brock, R. E. Glor, andW. Challenger.2008. GEIGER: investigating evolutionary radiations. Bioinfor-matics 24:129–131.

Harmon, L. J., J. B. Losos, T. J. Davies, R. G. Gillespie, J. L. Gittleman,W. B. Jennings, K. H. Kozak, M. A. McPeek, F. Moreno-Roark,

T. J. Near, A. Purvis, R. E. Ricklefs, D. Schluter, J. A. Schulte II,O. Seehausen, B. L. Sidlauskas, O. Torres-Carvajal, J. T. Weir, andA. Ø. Mooers. 2010. Early bursts of body size and shape evolutionare rare in comparative data. Evolution 64:2385–2396.

Hu, S. X., Q. Y. Zhang, Z. Q. Chen, C. Y. Zhou, T. Lü, T. Xie,W. Wen, J. Y. Huang, and M. J. Benton. 2011. The Luoping biota:exceptional preservation, and new evidence on the Triassicrecovery from end-Permian mass extinction. Proceedings of theRoyal Society of London B 278:2274–2282.

Hudson, J. D., D. M. Martill, and K. N. Page. 1991. Fossils of theOxford Clay. Palaeontological Association Field Guide to Fossils4. Wiley, London.

Hughes, M., S. Gerber, and M. A. Wills. 2013. Clades reach highestmorphological disparity early in their evolution. Proceedings ofthe National Academy of Sciences USA 110:13875–13879.

Ilin, A., and T. Raiko. 2010. Practical approaches to principal com-ponent analysis in the presence of missing values. Journal ofMachine Learning Research 11:1957–2000.

Ingram, T., L. J. Harmon, and J. B. Shurin. 2012. When should weexpect early bursts of trait evolution in comparative data? Pre-dictions from an evolutionary food web model. Journal ofEvolutionary Biology 25:1902–1910.

Kelley, N. P., and N. D. Pyenson. 2015. Evolutionary innovationand ecology in marine tetrapods from the Triassic to theAnthropocene. Science 348:aaa3716.

Kelley, N. P., R. Motani, D.-Y. Jiang, O. Rieppel, and L. Schmitz. 2014.Selective extinction of Triassic marine reptiles during long-term sealevel changes illuminated by seawater strontium isotopes. Palaeo-geography, Palaeoclimatology, Palaeoecology 400:9–16.

Korn, D., M. J. Hopkins, and S. A. Walton. 2013. Extinction space—amethod for the quantification and classification of changesin morphospace across extinction boundaries. Evolution 67:2795–2810.

Laurin, M. 2004. Evolution of body size, Cope’s rule and the originof amniotes. Systematic Biology 53:594–622.

Liu, J., S. X. Hu, O. Rieppel, D.-Y. Jiang, M. J. Benton, N. P. Kelley,J. C. Aitchison, C. Zhou, W. Wen, J. Huang, T. Xie, and T. Lv.2014. A gigantic nothosaur (Reptilia: Sauropterygia) from theMiddle Triassic of SW China and its implication for the Triassicbiotic recovery. Scientific Reports 4:7142.

Massare, J. A. 1987. Tooth morphology and prey preference ofMesozoic marine reptiles. Journal of Vertebrate Paleontology7:121–137.

Motani, R. 1997. Redescription of the dentition ofGrippia longirostris(Ichthyosauria) with a comparison with Utatsusaurus hataii.Journal of Vertebrate Paleontology 17:39–44.

——. 2005. Evolution of fish-shaped reptiles (Reptilia:Ichthyopterygia) in their physical environments and constraints.Annual Review of Earth and Planetary Sciences 33:395–420.

Motani, R., X. H. Chen, D.-Y. Jiang, L. Cheng, A. Tintori, andO. Rieppel. 2015a. Lunge feeding in early marine reptiles andfast evolution of marine tetrapod feeding guilds. Scientific Reports5:8900.

Motani, R., D.-Y. Jiang, G.-B. Chen, A. Tintori, O. Rieppel, C. Ji, andJ. D. Huang. 2015b. A basal ichthyosauriform with a short snoutfrom the Lower Triassic of China. Nature 517:485–488.

Navarro, N. 2003. MDA: a Matlab-based program for morphospace-disparity analysis. Computers and Geosciences 29:655–664.

Neenan, J. M., C. Li, O. Rieppel, F. Bernardini, C. Tuniz, G. Muscio,and T. M Scheyer. 2014. Unique method of tooth replacement indurophagous placodont marine reptiles, with new data on thedentition of Chinese taxa. Journal of Anatomy 224:603–613.

Nordén, K. K., C. J. Duffin, and M. J. Benton. 2015. A marinevertebrate fauna from the Late Triassic of Somerset, and a reviewof British placodonts. Proceedings of the Geologists’Association.doi:10.1016/j.pgeola.2015.07.001.


Peters, S. E. 2005. Geologic constraints on the macroevolutionaryhistory of marine animals. Proceedings of the National Academyof Sciences USA 102:12326–12331.

Polcyn, M. J., L. L. Jacobs, R. Araújo, A. S. Schulp, and O. Mateus.2014. Physical drivers of mosasaur evolution. Palaeogeography,Palaeoclimatology, Palaeoecology 400:17–27.

Pyenson, N. D., N. P. Kelley, and J. F. Parham. 2014. Marine tetra-pod macroevolution: physical and biological drivers on 250 Maof invasions and evolution in ocean ecosystems. Palaeogeo-graphy, Palaeoclimatology, Palaeoecology 400:1–8.

R Development Core Team. 2011. R: a language and environmentfor statistical computing. R Foundation for Statistical Comput-ing, Vienna, Austria.

Rieppel, O. 1998. Corosaurus alcovensis Case and the phylogeneticinterrelationships of Triassic stem-group Sauropterygia.Zoological Journal of the Linnean Society 124:1–41.

——. 2002. Feeding mechanics in Triassic stem-group sauropter-ygians: the anatomy of a successful invasion of Mesozoic seas.Zoological Journal of the Linnean Society 135:33–63.

Ross, M. R. 2009. Charting the Late Cretaceous seas: mosasaurrichness and morphological diversification. Journal of VertebratePaleontology 29:409–416.

Royall, R. M. 1997. Statistical evidence: a likelihood paradigm.Chapman and Hall, New York.

Ruta, M., K. D. Angielczyk, J. Fröbisch, and M. J. Benton. 2013.Decoupling of morphological disparity and taxic diversityduring the adaptive radiation of anomodont therapsids.Proceedings of the Royal Society of London B 280:20131071.

Sallan, L. C., and M. Friedman. 2012. Heads or tails: stageddiversification in vertebrate evolutionary radiations. Proceedingsof the Royal Society of London B 279:2025–2032.

Scheyer, T. M., C. Romano, J. Jenks, and H. Bucher. 2014. EarlyTriassic marine biotic recovery: the predators’ perspective. PLoSONE 9:e88987.

Schluter, D. 2000. The ecology of adaptive radiation. OxfordUniversity Press, Oxford.

——. 2001. Ecology and the origin of species. Trends in Ecology andEvolution 16:372–380.

Schneider, C. A., W. S. Rasband, and K. W. Eliceiri. 2012. NIHImage to ImageJ: 25 years of image analysis. Nature Methods9:671–675.

Schwimmer, D. R., J. D. Stewart, and G. D. Williams. 1997. Xiphacti-nus vetus and the distribution of Xiphactinus species in the easternUnited States. Journal of Vertebrate Paleontology 17:610–615.

Simpson, G. G. 1944. Tempo and mode in evolution. ColumbiaUniversity Press, New York.

——. 1953. The major features of evolution. Columbia UniversityPress, New York.

Slater, G. J., S. A. Price, F. Santini, and M. E. Alfaro. 2010. Diversityversus disparity and the radiation of modern cetaceans.Proceedings of the Royal Society of London B 277:3097–3104.

Smith, A. B., A. S. Gale, and N. E. A. Monks. 2001. Sea level changeand rock-record bias in the Cretaceous: a problem for extinctionand biodiversity studies. Paleobiology 27:241–253.

Smithwick, F. M. 2015. Feeding ecology of the deep-bodied fishDapedium (Actinopterygii, Neopterygii) from the Sinemurian ofDorset, England. Palaeontology 58:293–311.

Steeman, M. E., M. B. Hebsgaard, R. E. Fordyce, S. Y. Ho,D. L. Rabosky, R. Nielsen, C. Rahbek, H. Glenner, M. V. Sørensen,and E. Willerslev. 2009. Radiation of extant cetaceans driven byrestructuring of the oceans. Systematic Biology 58:573–585.

Stubbs, T. L., S. E. Pierce, E. J. Rayfield, and P. S. L. Anderson. 2013.Morphological and biomechanical disparity of crocodile-linearchosaurs following the end-Triassic extinction. Proceedings ofthe Royal Society of London B 280:20131940.

Taylor, M. A. 1987. How tetrapods feed in water: a functionalanalysis by paradigm. Zoological Journal of the Linnean Society91:171–195.

Thorne, P. M., M. Ruta, and M. J. Benton. 2011. Resetting theevolution of marine reptiles at the Triassic–Jurassic boundary.Proceedings of the National Academy of Sciences USA 108:8339–8344.

Vermeij, G. J. 1977. The Mesozoic marine revolution: evidence fromsnails, predators and grazers. Paleobiology 3:245–258.

Wellborn, G. A., and R. B. Langerhans. 2015. Ecological opportu-nity and the adaptive diversification of lineages. Ecology andEvolution 5:176–195.

Yoder, J. B., E. Clancey, S. Des Roches, J. M. Eastman, L. Gentry,W. Godsoe, T. J. Hagey, D. Jochimsen, B. P. Oswald, J. Robertson,B. A. J. Sarver, J. J. Schenk, S. F. Spear, and J. L. Harmon. 2010.Ecological opportunity and the origin of adaptive radiations.Journal of Evolutionary Biology 23:1581–1596.

Young, M. T., S. L. Brusatte, M. Ruta, and M. B. De Andrade. 2010.The evolution of Metriorhynchoidea (Mesoeucrocodylia,Thalattosuchia): an integrated approach using geometricmorphometrics, analysis of disparity, and biomechanics.Zoological Journal of the Linnean Society 158:801–859.


stubbs, t. l. , & benton, m. j. (2016). ecomorphological ... · thomas l. stubbs and michael j....

Documents