Transactionsof the Royal Society of Edinburgh: Earth Sciences, 92, 229±261, 2002 (for 2001)

Thetetrapod Caerorhachis bairdi Holmes and Carroll from the Lower Carboniferous of Scotland MarcelloRuta, Andrew R.Milnerand Michael I. Coates

ABSTRACT:The tetrapod Caerorhachisbairdi ,probablyfrom the Pendleian Limestone Coal Groupin the Scottish Midland Valley, is rediagnosed and redescribed, and its a nities are discussed. Caerorachis wasoriginally interpreted as a temnospondylamphibian, based on charactersthat are now regarded as primitive for tetrapods, or of uncertain polarity. Several features of Caerorhachis (e.g.gastrocentrous vertebrae, curved trunk ribs, reduced dorsal iliac blade,L-shaped tarsal intermedium) are observed in certain primitive amniotes. In particular, Caerorhachis resembles ‘anthracosaurs’,generally considered to be among the most primitive of stem-groupamniotes. Thephylogenetic position of Caerorhachis isconsidered in the light of recentlypublished cladistic analysesof Palaeozoictetrapods. Most analyses place Caerorhachis atthe base of, or within, ‘anthra- cosaurs’.When multiple, equally parsimonious solutions are found, its ‘anthracosaur’a nities are shownin at least some trees, and are supported by several informative and, generally, highly consistentcharacters. Alternative phylogenetic placements (e.g. sister taxon to temnospondyls) are usuallyless well corroborated. Ifthefundamental evolutionary split of mostearly tetrapods into stem-group lissamphibians (e.g. temnospondyls)and stem-group amniotes (e.g. ‘anthracosaurs’)is accepted, then the revised interpretationof Caerprhachis shedslight on near-ancestralconditions for Amniota.

KEYWORDS:amniotes, ‘anthracosaurs’,characters, lower jaw, palate, relationships, skull.

Inthe present work, we reconstructand redescribe Caerorha- nostrilsare similar to those of baphetids(=loxommatids), an chis bairdi,oneof theproblematic fossil tetrapods ‘...fall[ing] enigmaticclade of large Carboniferous tetrapods with key- intothe nebulous area which surrounds the crown group node’ hole-shapedorbits (Beaumont 1977; Beaumont & Smithson (Ahlberg1998, p. 104). Originally described by Holmes & 1998;Milner & Lindsay1998), variously regarded as stem- Carroll(1977), Caerorhachis wasplaced in the temnospondyls, grouptetrapods (e.g. Milner et al.1986;Ahlberg & Milner adiversegroup of LowerCarboniferous to LowerCretaceous 1994)or stem-group amniotes (e.g. Panchen & Smithson tetrapodsoften considered to be ancestralto someor alllissam- 1988;Clack 1998a,b, in press). More recently, Coates (1996) phibians(e.g. Bolt 1979, 1991; Milner 1988, 1990; Milner & hypothesisedthat Caerorhachis mayrepresent a basalplesion Sequeira1994). Several classi® cations of earlytetrapods place inthe amniote stem-group, as indicated by its gastrocentrous Caerorhachis eitherin a basalposition within temnospondyls, vertebrae,transverse pelvic ridge, and L-shaped tarsal interme- oras their plesiomorphic sister-taxon, sometimes in a mono- dium.Indeed, various cranial and postcranial characters of typicfamily Caerorhachidae (e.g. Godfrey et al.1987;Carroll Caerorhachis resemblemorphological conditions found in 1998;Milner 1988, 1990, 1993a). This interpretation has been severalmembers of the ‘anthracosaurs’(e.g. Holmes 1984, acceptedby most workers until recently (e.g. Holmes 2000). 1989;Panchen 1985; Smithson 1986, 1994; Clack 1987a,b, However,some of the characters employed by Holmes & 1994,1998a,b, 2001; Panchen & Smithson1987, 1988; Smith- Carroll(1977) to assign Caerorhachis totemnospondyls are son et al.1994;Lebedev & Coates1995; Coates 1996; Paton broadlydistributed among Palaeozoic tetrapods (e.g. radiating et al. 1999). ‘Anthracosaurs’,used throughout sensu Panchen dermalsculpture), whereas others are of dubious polarity (e.g. &Smithson(1988), include Permo-Carboniferous tetrapods rigidcontact between skull roof and cheek) (Milner & Sequiera placedby various authors among the most basal of stem- 1994;Coates 1996; sections 5.1 and 5.2). A. R.Milner(1980) groupamniotes (e.g. embolomeres and gephyrostegids; see ®rstsuggested that hardly any anatomical feature of Caerorha- alsoSa Ève-SoÈderbergh1934; Carroll 1970; Heaton 1980; Smith- chis canbe regarded as asynapomorphyof this tetrapod with son1986, 1994; Gauthier et al.1988;Clack 1994, 1988a,b; temnospondyls,apart from its rather small and narrow inter- Smithson et al.1994;Coates 1996; Paton et al.1999).In con- pterygoidvacuities. In particular, the presence of large and trast,Laurin’s (1998a± c), Laurin & Reisz’s(1997, 1999), and horseshoe-shapedpleurocentra distinguishes Caerorhachis Laurin et al.’s(2000a,b) novel taxonomic use restricts the fromeven the most primitive temnospondyls (Milner & term ‘anthracosaurs’to a cladeincluding Solenodonsaurus , Sequeira1994, 1998). Milner & Sequeira(1994) discussed diadectomorphsand crown-group amniotes (but see comments Caerorhachis inthe context of ananalysisof characterdistribu- in Coates et al. 2000). tionin basal temnospondyls, and concluded that it should be Givensuch contrasting views on the a nities of Caerorha- removedfrom this group. According to Milner & Sequeira chis,aredescriptionis necessary in order to provide a more (1994), Caerorhachis isbest regarded as atetrapodof uncertain sharplyde® ned picture of this taxon. Re-examination of anities. However, they noted that its palate and internal Caerorhachis bringsinto focus some additional characters 230 MARCELLORUTA ET AL. thatcomplement its original description (Holmes & Carroll toa paraphyletic ‘anthracosaurs’.This placement may re¯ ect 1977),and prompts a revisedinterpretation of several other eitherpaucity of key features linking Caerorhachis unambigu- characters.Recent advances in our understanding of early ouslyto other taxa, or the primitive nature of its lower jaws. tetrapodcomparative anatomy reveal the importance of this Similarresults are obtained by analysing Laurin’s (1998a) fossil:as in the case of the recently described Eucritta (Clack matrix.However, manipulations of Ahlberg & Clack’s(1998) 1998b,2001), Caerorhachis showsan interesting combination andLaurin’s (1998a) data sets, based on taxon and/ orcharacter ofplesiomorphiesand apomorphies that may explain the con- inclusion,place Caerorhachis within ‘anthracosaurs’(see ¯ictsurrounding its interpretation. Therefore, a reappraisalof ‘Anities’ below). thisanimal may cast new light on morphological conditions TheAppendix includes the coding for Caerorhachis relative occurringat the evolutionary split between di Verentlineages toseven recent analyses (Carroll 1995; Coates 1996; Ahlberg ofearly tetrapods. This study examines only some aspects of &Clack1998; Clack 1998a,b; Laurin 1998a; Paton et al. theanatomy of Caerorhachis .Comparisonswith other early 1999;see Appendix). However, only four of theseare discussed, tetrapodspermit a revisedinterpretation of suture patterns namelyCoates (1996), Ahlberg & Clack(1998), Laurin (1998a), andbone proportions on the skull roof and palate. A thorough and Paton et al.(1999).These were chosen according to two redescriptionof the lower jaw is provided in the light of criteria.First, Laurin’s (1998a) and Paton et al.’s(1999) data Ahlberg& Clack’s(1998) recent work on mandibular charac- setssupersede Carroll’s (1995) and Clack’s (1998a,b), respec- tersin early tetrapods. Several characters of the palate, man- tively,in terms of character inclusion and taxon sampling. dibleand postcranial skeleton that di Verfrom those of Therefore,they are likely to provide a morestringent test of temnospondylsare discussed in detail (e.g. spindle-shaped thea nities of Caerorhachis .Second,Ahlberg & Clack’s palatalvacuities, morphology of choanae, overall con® guration (1998)and Laurin’s (1998a) works o Vera novelperspective ofmandibular ramus, shape of basal plate of parasphenoid, onPalaeozoictetrapod phylogeny, by placingmost groups on gastrocentrousvertebrae, transverse pelvic ridge, etc.). New thetetrapod stem. This perspective contrasts with the hypo- featuresof the pelvic girdle and hind-limb are highlighted thesiseddeep evolutionary split of most Palaeozoic tetrapods (e.g.short dorsal iliac blade, pelvic sutures, acetabular shape, intostem-group lissamphibians (e.g. temnospondyls) and ischiotrochanterictuberosity on femur, details of ¯exorsurface stem-groupamniotes (e.g. ‘anthracosaurs’),as advocated by oftibia,details of tarsalelements, etc.). Some characters of the Coates(1996) and Paton et al.(1999).Investigation of con¯ict- axialskeleton and body armour not discussed by Holmes & ingsolutions produced by di Verentmatrices falls outside the Carroll(1977) are mentioned brie¯ y (e.g.haemal spine mor- aimsof this paper, and will bediscussed elsewhere in conjunc- phology,shape of gastralia, etc.). The atlas, axis, fore-limb tionwith a newlarge-scale matrix for early tetrapods. andpectoral girdle do not require a redescription.The new reconstructionof Caerorhachis (Fig. 1A) diVers in some respectsfrom Holmes & Carroll’s(1977) (Fig. 1B). In particu- lar,conservative estimates of lower jaw and presacral column 1.Materialsand methods lengthsindicate that the head is more massive than previously assumed.Stout shape and relatively small size of the femur Caerorhachis ispreserved as acid-etchedmoulds of cranialand (comparedwith mandible length), and the very short tibia postcranialmaterial, probably belonging to asingleindividual and® bula,contribute to the diminutive aspect of the hind- (Holmes& Carroll1977). Silastomer casts of di Verentparts of limbs.Incomplete remains of the fore-limb suggest that the theskeleton were measured (Table 1) anddrawn using a Zeiss latter,too, must have been rather small. The length ratios of binocularmicroscope equipped with a cameralucida, and head,trunk and tail are approximately 1 : 2 : 1 . The total photographedunder low-angle cross-lighting to enhancerelief. bodylength must have exceeded 200 mm. The overall ‡ propor- Consultedmatrices were re-built in MacClade 3.0.5 tionsare reminiscent of those of certain Australasian skinks, (Maddison& Maddison1992), and re-processed with PAUP* e.g.representatives of the genus Tiliqua (Pough et al. 2000). 4.0b4a (SwoVord2000) on a PowerMacG4 under original Inthe analyticalpart of this paper, several recently published weightingand ordering regimes. Search settings were as follows datamatrices are used to test di Verenthypotheses about the (see SwoVord1993 for explanations): heuristic random step- a nities of Caerorhachis withwidely recognised and (mostly) wise additionsequence (100 replicates; 10 trees held at each well-characterisedPalaeozoic groups. The rationale underlying step);ACCTRAN optimisation; TBR algorithm;MULPARS ourapproach is that the systematic limits and content of major andSTEEPEST DESCENT optionsin e Vect.Basic tree statis- Palaeozoicgroups are, with some exceptions, relatively simple ticsexcluded uninformative characters. Changes in character tode® ne (e.g. Panchen 1980; Smithson 1982; Hook 1983; weightingand/ orordering were considered in some analyses, Panchen& Smithson1987, 1988; Milner 1988, 1990, 1993a,b; andthe resulting cladograms were compared with those of Carroll1995; Carroll et al.1998;Milner & Lindsay1998). theoriginal parsimony runs. With multiple solutions, we Therefore,it ispossible to assess the characters of Caerorhachis summarisedcladogram information using strict consensus againstthose considered to be apomorphous for various trees.Reweighting of character used the maximum (best ® t) groups,regardless of the mutual placements of the latter in valueof their rescaled consistency indexes. Branch support diVerentcladograms. Con¯ icting phylogenetic positions for wasevaluated using bootstrap and decay index. Bootstrap Caerorhachis (e.g.ambiguous placement in equally parsimo- analyses(1000 replicates; groups compatible with 50% nioustrees or in trees deriving from di Verentdatabases) can majority-ruleconsensus), were performed via simple addition thenbe discussed in the light of reconstructed patterns of sequence(10 trees held at eachstep). However, time-consuming characterdistribution. Although not all the consulted data- processingof Ahlberg & Clack’s(1998) and Laurin’s (1998a) basesprovide unambiguous results with regard to the system- matricesforced us to employ a ‘fast’stepwise addition aticinterpretation of this animal (see ‘Anities’ below), the option.For the same reason, decay values could not be calcu- ‘anthracosaur’a nities indicated by somestudies (e.g. Coates latedfor those two matrices. Only scorable characters support- 1996; Paton et al.1999)are linked to characters that are usually ingthe position of Caerorhachis wereconsidered. A detailed moreconsistent and, often, more numerous than those favour- breakdownof each analysis, including performance of all ingalternative phylogenetic placements. In thecase of Ahlberg characterstate changes across all cladograms, is available &Clack’s(1998) lower jaw data set, Caerorhachis appearsnext uponrequest. CAERORHACHIS BAIRDI 231

Figure 1 Caerorhachisbairdi Holmes& Carroll: (a)revised reconstruction, and (b) Holmes & Carroll’s recon- structionof the skeleton, left lateral aspect and to the same length; (c) revised reconstruction, and (d) Holmes &Carroll’s reconstructionof skull, lateraland dorsal aspects and to the same length; scale bars 1 cm. ˆ 232 MARCELLORUTA ET AL. Table 1 Anatomicalmeasurements of Caerorhachis in mm.

2.Systematicpalaeontology fromfemoral head by trough-like space; interepipodial space betweentibia and ® bula. TETRAPODAGoodrich, 1930 Leastambiguous synapomorphies of Caerorhachis with stem- Class,Order and Family unassigned groupamniotes: subcirculardepression on parasphenoid Genus Caerorhachis Holmes& Carroll,1977 delimitedby pronounced peripheral rim; enlarged U-shaped pleurocentra;reduced dorsal iliac blade; L-shaped tarsal inter- Typespecies. Caerorhachisbairdi Holmes& Carroll,1977. medium. Diagnosis. Asfor the only species. Possibleadditional synapomorphies of Caerorhachis with stem- groupamniotes: rearmostpart of mesial lamina of splenial Caerorhachisbairdi Holmes& Carroll,1977 closerto anterior margin of adductor fossa than to lower jaw (Figs 1±15) symphysis;cervical ribs straight, morphologically distinct fromtrunk ribs and with ¯ atteneddistal ends; curved trunk Holotypeand only known specimen. MCZ 2271;Museum of ribs;transverse ridge sweeping from lateral surface to ventral ComparativeZoology, Harvard University. Incomplete indi- marginof posterioriliac process. vidualon associated blocks of coal shale, sold to Harvard Possiblesynapomorphies of Caerorhachis withbaphetids: elon- Universityin 1883 as part of the Thomas Stock collection of gatechoana, slightly wider posteriorly than anteriorly, close ScottishCarboniferous vertebrates. MCZ 2271has no accom- toupper jaw margin, and with postero-mesial angle at anterior panyinggeographical or stratigraphical information, hence endof vomer/ palatinesuture; ?poorly pronounced lateral out- theuse of question marks at the beginning of the following growthof prefrontal. sections.According to Holmes& Carroll(1977) and Smithson 3. Symplesiomorphies :intertemporal;supratemporal/ post- (1985a),the shale lithology is similar to that of the Loanhead parietalsuture; anteroposteriorly elongate subtemporal fossa; No.2 Ironstone,from which other specimens in the Stock vomerand palatine fangs; radiating dermal ornament; large collectionhave been recorded. numberof dentary teeth; parasymphysial plate with two Locality. ?Ramsaycolliery, Loanhead (Edinburgh), Mid- fangsand scattered denticles; denticulated parasphenoid; lothian,Scotland. uniformlydenticulated palate; mobile basal articulation; Horizon. ?UpperLimestone Coal Group; Loanhead No. 2 crescent-shapedintercentra; length of trunk ribs less than Ironstone(Holmes & Carroll1977); non Rumbles(Burghlee) twicethe combined height of pleurocentrumand neural spine; Ironstone(Smithson 1985a), ®de Andrews& Brand(1991). dorsaliliac blade and posterodorsal iliac process; incipient Age. ?Baseof Serpukhovian,Pendleian Stage, Cravenoceras distalreduction of femoral adductor blade; pentadactyl pes; Zone(Harland et al. 1989). gastralia. Diagnosis (inthe context of post-Devonian tetrapods; see 4. Othercharacters of uncertainpolarity :palatine/ectoptery- also ‘Remarks’below). goidfangs 3± 4 timeslarger than vomerine fangs; high head/ trunklength ratio; high femur/ puboischiumlength ratio. 1. Presumedautapomorphies :subtrapezoidaltabulars with majoraxis at about 45 tosagittal axis; toothless maxillary regionat level of palatine fangs; subpentagonal choana; elongateanterodorsal process of posterior coronoid; densely 3.Description denticulatedmesial surface of lower jaw. 2. Synapomorphies (atvarious levels of taxonomicinclusive- Thematerial consists of a weatheredskull roof partly preserved ness):broadly crescent-shaped postorbitals, slightly elongate asanendocast; slightly disrupted palate; incomplete basal plate anteroposteriorly,and about as large as supratemporals; andremains of cultriform process of parasphenoid; almost narrow,spindle-shaped interpterygoid vacuities, less than half completemaxillae in ventral and mesial views; ¯ attened,but aswide as skull; absence of marginal teeth on ectopterygoids, virtuallycomplete mandibular rami; partially articulated ver- palatinesand vomers; rearward extension of parasphenoid; tebralcolumn in left aspect, including some cervical, trunk dorsomesialorientation of mandibularadductor fossa; absence andcaudal ribs; proximal tail region with haemal spines; ofcoronoid fangs and tooth rows; mesial laminae of infra- crushedleft half of pelvic girdle; left femur in extensor view; dentaries;small surangular crest; posterior coronoid with lefttibia and incomplete left ® bulain ¯ exorview; disarticulated small,posterodorsal projection entering adductor fossa; fully leftfoot; incomplete interclavicle and plate of right clavicle; ossi®ed olecranon; boss-like internal trochanter separated fragmentsof fore-limb;gastralia. CAERORHACHIS BAIRDI 233

Figure 2 Caerorhachisbairdi Holmes& Carroll: (a)MCZ 2271,incomplete skull tableand endocast of preorbital region;(b) camera lucida drawing; scale bars 1 cm. ˆ 234 MARCELLORUTA ET AL. 3.1Skull table connectionwith the preorbital region. Incomplete endocasts Theincomplete skull table (Fig. 1C, D) isdisrupted and mostly ofthefrontals are visible anterior to suchdigitations, together representedby an endocast (Fig. 2A). A posteriorsmall area witha weaktrace of theinterfrontal suture. withdermal sculpture is visible. Holmes & Carroll’s(1977) Thesubtrapezoidal postparietals resemble those of Crassi- interpretationof thelatter is followed here, but we re-interpret gyrinus (Panchen1985; Clack 1998a). Their posterior margins apoorlypreserved endocast surface immediately to the right of areslightly convex. However, we areuncertain about the occur- thesculptured area (Figs 1C, 2B). According to Holmes & renceof postparietal projections. These would correspond to Carroll(1977), the surface in question includes the right supra- smallsubtriangular areas near the posteromesial angle of temporal,tabular and squamosal (Fig. 1D). However, we eachpostparietal in Holmes & Carroll’s(1977) reconstruction, suggestthat the bone identi® ed by them as the right supra- buttheir size may have been exaggerated by ¯ attening. temporalbelongs in fact to the rightmost part of the right Theelongate, vaguely subtrapezoidal supratemporal endo- parietal.Anterolateral to this lies a smallsubtriangular area castsappear almost featureless, and their margins are scarcely thatmay correspond to a weatheredright intertemporal. visible.Most of the left supratemporal is overlapped by two Afragmentof endocast, possibly belonging to the right post- endocastfragments visible anterolateral to the left parietal. frontal,is visible anteromesial to the right intertemporal. Thelarger fragment is undoubtedly a postorbital,based on its Thearea of endocast interpreted by Holmes & Carroll(1977) crescent-likeshape and concave anterior margin. The smaller asthe right tabular is probably a continuationof the right fragment,partly covering the postorbital, is likely to correspond postparietal,whereas the area corresponding to their right toan anteromesiallydisplaced left intertemporal. Its posterior squamosalis considered to belong to the right supratemporal. two-thirdsis approximately triangular and plectrum-shaped. Theirregular suture between the right parietal and the right Itspoorly preserved anterior margin is gentlyconcave. supratemporalcontacts the anterolateral angle of the right Afragmentof endocastlying just anterior to theanterolat- postparietal.The course of the supratemporal/ postparietal eralangle of the left parietal, a shortdistance away from the sutureis poorly preserved. anteriormargin of the left postorbital, deserves some com- Althoughthe suture pattern at theback of theskull roof is ments.Its outline is poorly de® ned. It is dividedinto a broadly temnospondyl-like(possibly primitive for tetrapods; Lombard squareposterior region, and an elongate, triangular region &Bolt1995; Clack 1998a,b; Paton et al.1999),it is important taperingrapidly anteriorly. Its overall shape suggests that it tonotethat the supratemporal/ postparietalcontact may re¯ect, mightrepresent either a postfrontalor a prefrontal,based on atleastin part, the con® guration of the ventral side of theskull comparisonswith such tetrapods as Balanerpeton or Eucritta roofonly. The spatial relationships of the supratemporal and (Milner&. Sequeira1994; Clack 1998b, 2001). It is di cult postparietalon the ventral side do not always conform to the toascertain whether this bone is preserved in its original orien- patternobserved on thedorsal side, as Smithson (1986) demon- tationor has been rotated through 180 .Ifit is indeed a pre- stratedin thecase of Proterogyrinuspancheni .Inthis tetrapod, frontalpreserved approximately in its original anatomical thedorsal side of the skull roof shows the typical tabular/ orientation,then the circumorbital series may reveal a pattern parietalcontact found in all ‘anthracosaurs’.Ventrally, these similarto that observed in baphetids (Beaumont 1977; Beau- twobones are separated more or lesscompletely by anarrow, mont& Smithson1998; Milner & Lindsay1998). With the strip-like,and anterolateral projection of the postparietal. In exceptionof Spathicephalus ,allbaphetids show pronounced addition,the separation between tabular and parietal is com- outgrowthsof theprefrontals and postorbitals or jugals. Such pleteon the right half of the ventral side, where the antero- outgrowthsdelimit the boundary between orbits and antorbital lateralprojection of the postparietal forms a smallsuture vacuities,resulting in a characteristic,keyhole-shaped constric- withthe supratemporal. On the left half, however, tabular tionof the combined orbit/ antorbitalvacuity. Assuming the andparietal retain a pointcontact. In Proterogyrinusscheelei correctidenti® cation and orientation of the prefrontal, the (Holmes1984) and Archeriacrassidisca (Holmes1989), the circumorbitalbone pattern suggests the intriguing possibility spatialrelationships between tabular and parietal are identical that Caerorhachis possessedbaphetid-like angular orbits. onbothsides of the skull roof, and this con® guration is found Giventhe poor preservation, this suggestion is extremely tenta- alsoin other ‘anthracosaurs’. tiveand in our reconstruction (Fig. 1C), we havedrawn more Thecrushed right tabular of Caerorhachis isvisible posterior conventional,subelliptical orbits. Angular orbits, though not tothe right supratemporal, from which it is delimited by a necessarilykeyhole-shaped, are known in several other tetra- stronglysinuous and interdigitating suture, interpreted by pods,including Anthracosaurus (Panchen1977; Clack 1987a), Holmes& Carroll(1977) as part of the posterior skull table Crassigyrinus (Panchen1973, 1985; Clack 1998a), and Whatch- margin.The tabular is subtrapezoidal, slightly larger than the eeria (Lombard& Bolt1995). intertemporal,and with a bluntposterior margin projecting Thepresence or absence of otic embayments cannot be slightlyposterior to the postparietals. Its main axis is oblique securelydetermined. The lateral margin of the tabular may tothe sagittal axis. Unfortunately, its relationships with the havecontributed to a veryshallow notch together with the pos- postparietalare partly obscured by the poor preservation of teriormargin of thesupratemporal, as in Loxomma or Eucritta itsmesial portion. (Beaumont1977; Clack 1998b, 2001), but an accurate recon- Thesubrectangular parietals are almost twice as long as structionof the temporal region is not possible, since both wide.Deep left and right indentations along the course of the thetabular and the posterior part of the supratemporal are irregularlysinuous interparietal suture indicate the probable heavilyeroded. positionof thepineal foramen, although the latter is not clearly Thepostorbital and the incomplete quadratojugal and visible.In the absence of additionalmaterial, it is impossible to squamosal(Fig. 3A, B) weredescribed by Holmes & Carroll ascertainwhether this condition re¯ ects preservation or (1977),and will notbe examined further. An endocast of the whetherthe pineal foramen underwent obliteration by ossi®ca- preorbitalskull table is associated with the disrupted left half tionduring ontogeny, as in medium- to large-sizedspecimens of ofthe palate (Fig. 5A), but details of the bones are obscure. thetemnospondyls Cochleosaurus and Chenoprosopus (Milner Abroadnasal with an interdigitating anterior margin and &Sequeira1998). Anterior to thesculptured area, the parietals shallowventral grooves is visible distally (Fig. 5A). Similar arerepresented by partially weathered endocasts. Their grooveswere observed by Clack(1987b) on theinternal surface anteriormargins show irregular digitations, indicating rigid ofthelacrimals and nasals in Pholiderpeton . CAERORHACHIS BAIRDI 235

Figure 3 Caerorhachisbairdi Holmes& Carroll: (a)MCZ 2271,mesial aspect of left maxilla, left postorbital in dorsalview; (b) MCZ 2271,incomplete left cheek and skull roof, disruptedendocast of preorbital region, mesialaspect of left maxilla, incomplete and disarticulated left half of palate, almost complete basal plate of para- sphenoid,partially preserved atlas and axis; (c) camera lucida drawing of anteriorend of left maxilla, showing funnel-shapedexcavation (posterolateral wall ofthechoana); scale bars 1cmfor (a,b); 0 :5 cm for (c). ˆ 236 MARCELLORUTA ET AL.

Figure 4 Caerorhachisbairdi Holmes& Carroll: (a)MCZ 2271,incomplete right half of palatearticulated with rightmaxilla, mesial aspect of rightlower jawramus; (b) camera lucida drawing of right half of palate and right maxilla;scale bars 1 cm. ˆ

3.2.Maxilla thickening,and ventrally by the anterior continuation of the Themaxilla (Figs 3A, 4, 5A), deeper in its anterior half than mesialshelf The maxilla shows about 45 tooth positions. The inits posterior half, carries a poorlydeveloped mesial shelf maxillaryteeth are comparable in size with the dentary teeth. (Fig.5A). A gapin the tooth row, corresponding to thespace Thebasal three-quarters of their crowns is cylindrical. Some normallyoccupied by about four or ® vetooth positions, is maxillaryteeth show an abrupt anticlockwise twist of their visiblenear the palatine fangs (Holmes & Carroll1977). The striationsnear the crown base. rostralend of the left maxilla shows a funnel-shapedexcavation (posterolateralwall of thechoana; Figs 3A, C, 5A),delimited 3.3. Palate dorsallyby a sheetof bone with a distinct,rod-like mesial Combinedinformation from the two preserved portions of the CAERORHACHIS BAIRDI 237

Figure 5 Caerorhachisbairdi Holmes& Carroll: (a)camera lucida drawing of disarticulated left half of palate, left maxilla,basal plate of parasphenoid, and endocast of preorbital region; (b) Holmes & Carroll’s reconstructionof palate(anatomical right half); (c) revised reconstruction of palate(anatomical left half); denticle ® eldsomitted; scale bars 1 cm. ˆ palate(Figs 3B, 4, 5A) supports most of Holmes & Carroll’s prechoanalregion, as found in cochleosaurids (Milner & (1977)reconstruction. Minor di Verencesbetween our recon- Sequeira1998). The postchoanal extent of thevomers is uncer- struction(Fig. 5C) and theirs (Fig. 5B) concern the more tain,but an estimate of theposition of theirposterior margins anteriorlyplaced parasphenoid, the more oblique orientation indicatesthat these cannot have extended much posteriorly ofthe palatine/ ectopterygoidsuture, and the position of the withrespect to the level of the palatine fangs. If this was the anteriorextremity of the maxilla with respect to the antero- case,then the anterior ends of the palatine rami of the ptery- lateralangle of the vomer. Despite lack of informationconcern- goidswould display a unique,V-shaped con® guration, not ingthe shape and mesial extensions of the vomers and the knownin other tetrapods. Based on available evidence, it morphologyof the anteriormost part of the palatine rami of seemsmore likely that the pterygoids were wedged between thepterygoids, the interchoanal region appears to have been thevomers, and that the intervomerine suture did not show a ratherbroad, with contributions from both vomers and ptery- considerablebackward projection with respect to the position goids.The relative proportions of vomers, palatines and ofthechoanae. ectopterygoidsresemble those of baphetids (Beaumont 1977). Theinterpterygoid vacuities are less than half as wide as the Additionalbaphetid-like features include the overall shape skull(Figs 3B, 5A, C). Theydi Verfromthose of most temno- ofthe choana (Figs 3B, 4, 5C), broader posteriorly than spondylsin that they are not smoothly curved along their anteriorly,and the fact that the anterior end of the vomer/ anterior,lateral and posterior margins (Romer & Witter palatinesuture intersects the posteromesial angle of the 1942;Romer 1947; A. R.Milner1980; Milner & Sequeira1994; choana(Figs 3B, 4, 5C). The inferred degree of curvature of Schoch& Milner2000). The anterior half of the quadrate theanterior snout margin (see also Holmes & Carroll1977) ramiof the pterygoids (Figs 3B, 5A, C) isrelatively broad impliesthat the vomers are not likely to have shown an elongate andlacks a posterolateral¯ ange.Such a ¯angeconfers a 238 MARCELLORUTA ET AL. characteristicanterior constriction to the subtemporal fossa of anteroventralmargin with an irregularlysinuous shape. Close manytemnospondyls (Milner 1990; Yates & Warren2000), examinationreveals, however, that the anteroventral margin includingthe most primitive members of the group (Milner & ofthe fossa has been slightly folded and pushed inside the Sequeira1994, 1998). In addition, the palatine rami of the fossaitself, that its posterior margin has been slightly rotated pterygoidsare quite broad, rather than strap-like as inseveral dorsolaterally(away from the median plane), and that its basaland derived temnospondyls. dorsalmargin has been crushed. When such deformations are Backward-pointing,strongly recumbent and striated den- takeninto account, it is possible to restore in part the shape ticlesare uniformly distributed on thepalate. They are smaller ofthefossa. This is likely to havehad a roughlyelliptical out- (about 0:06mm diameter at the base) and more widely linein life, similar but not identical to that of the left fossa as spacedon theposterior part of thequadrate rami of the ptery- preserved.The latter appears distorted, albeit to a lesser goidsand near the anterolateral angles of the vomers. Small degreethan its antimere. Extreme ¯ atteninghas damaged its andirregular clusters of more robust denticles (about anteriormargin and squashed its postero-central third. The 0:17mmin diameter at the base) lie mesial and anterior to originalreconstruction of the fossa (Holmes & Carroll1977) thevomerine fangs. Similar clusters have been observed in wasbased largely on information from the left ramus. In the othertetrapods (e.g. Tulerpeton;Lebedev& Clack1993). The revisedreconstruction, the ventral margin of the fossa is not palataldenticles reach the rearmost end of the quadrate rami uniformlyconcave, but shows a slightbend more or less half- ofthepterygoids, as in Proterogyrinus (Holmes1984). Several wayalong its length. Such a bendis partly visible in the left denticlesappear broken or worn out. Details of the external ramus,where its curvature has been decreased by compaction, surfacesof thesmallest denticles are not discernible. However, andin the right ramus, where crushing has exaggerated its asfaras it can be determined,all denticles are striated, regard- curvature,and has resulted in the sharp, upward displacement lessof theirsize or position. An extensivelydenticulated palate ofitsposterior half. A moreor less prominent bend along the alsooccurs in Devonian taxa (Jarvik 1980, 1996), some micro- lowermargin of the fossa is also found (among others) in the saurs(Carroll & Gaskill1978), ‘anthracosaurs’(Carroll 1970; lowerjaws of Gephyrostegus and Eocaptorhinus (Carroll1970; Holmes1984), and several temnospondyls (Milner & Sequeira Heaton1979), as reconstructed by Ahlberg & Clack(1998), 1994;Holmes et al. 1998). and in Archeria,as® guredby Holmes (1989). However, in all thesetaxa the bend is situated posteriorly along the ventral marginof the fossa. Furthermore, Caerorhachis does not 3.4.Lower jaw possessa mesiallydirected ¯ angealong the postero-ventral Theessentially complete, albeit disrupted, lower jaw rami marginof the fossa, unlike Gephyrostegus,Archeria , some (Figs4A, 6) are better preserved mesially than laterally. Their temnospondyls,nectrideans and seymouriamorphs (Ahlberg lateralsurfaces, however, show the approximate position of &Clack1998). thecentres of radiation of dermal ornament on the dentary 3.4.3.Surangular crest. Theanterodorsal margin of the sur- andinfradentaries (Fig. 6A; see also Holmes & Carroll1977), angularforms a veryslight crest (Figs 4A, 6B, 7A). This region aswell asthe partially preserved course of some sutures, ofthelower jaw is intermediate morphologically between those especiallythose between dentary and surangular, and between of Megalocephalus and Crassigyrinus ,inwhich a surangular dentaryand splenial. The mesial aspect of the right ramus crestis absent, and those of Archeria,Pholiderpeton and most (Fig.4A) provides most of the available information on the temnospondyls,in which the crest forms a distinct,raised spatialrelationships of coronoids and infradentaries, and on step(Fig. 8). Caerorhachis also diVersfrom many ‘higher’ theshape of theadductor fossa. The left ramus (Fig. 6) shows temnospondyls( sensu Yates& Warren2000), in that the pos- thedentary and the symphysial region. teriorcoronoid is appressed, for a shortdistance backward, 3.4.1.General shape. Whenreconstructed in mesial view tothe mesial side of the surangular crest, and is not exposed (Fig.7A), the lower jaw ramus is rather deep in its posterior laterally,unlike the posterior coronoid of Pholiderpeton half,and tapers uniformly anteriorly. Its ventral margin is (Clack1987b), Archeria (Holmes1989), many temnospondyls moresmoothly curved than in Holmes & Carroll’s(1977) (Romer1947; Milner 1990; Yates & Warren2000), and several reconstruction(Fig. 7B). The ramus does not carry a retro- seymouriamorphs(White 1939; Bystrow 1944; Klembara 1997) articularprocess, and resembles in general proportions andamniotes s.s.(Romer1956). those of Crassigyrinus (Panchen1973, 1985), Megalocephalus 3.4.4.Backward extensions of dentaryand posterior coronoid. (Beaumont1977) and various ‘anthracosaurs’(Carroll 1970; Inlateral aspect, the left jaw ramus shows the backward extent Holmes1984, 1989; Clack 1987b, 1998a; Ahlberg & Clack ofthe dentary (Fig. 6A). In dorsal aspect, the rear end of the 1998)(see Fig. 8A± I, K±N forcomparisons with selected dentaryis subtriangular, and tapers rapidly to a point, lowerjaw rami of other extinct tetrapods). Jaw depth was contributingto the dorsal margin of the adductor fossa only calculatedfrom the reconstructed mesial aspect following the toa smalldegree. Conversely, in several primitive tetrapods methodoutlined by Clack (1987b). Depth was measured as (e.g. Acanthostega , Ichthyostega , Ventastega, Metaxygnathus , maximumlength of a lineperpendicular to one joining the Densignathus , Crassigyrinus ),thedentary shows a substantial anteriorend of the adductor fossa and the back of the articular, backwardextension (Fig. 8) (Ahlberg & Clack1998; Daeschler andexpressed as a percentageof total jaw length, the latter 2000).Immediately mesial to therearmost part of thedentary, wasmeasured along a linejoining the symphysis and the andseemingly applied closely to it,is an irregularprocess of the backof the articular. The depth/ lengthratio (about 20 :85%) posteriorcoronoid. The process projects further posteriorly isapproximately intermediate between those of Proterogyrinus alongthe anterodorsal margin of the adductor fossa, but less (19:2%) and Neopteroplax (22:35%)based on Clack’s(1987b) extensivelythan in Holmes & Carroll’s(1977) reconstruction. measurements. Therecumbent V-shaped posterior margin in their reconstruc- 3.4.2.Adductor fossa. Thedorsomesially facing adductor tionis likely to represent a fracturerunning across the dorsal fossa(Figs 4A, 6B, 7A) occupies about 26% of total jaw halfof the mesial surface of the surangular in the left ramus. length.Compaction and crushing of the fossa region are Theactual posterior margin of the process extends as far evidentin both rami. In the right ramus, the fossa appears to backas the anterior quarter of the fossa, and is rather beroughly triangular, with a distinctstraight posterior narrow(Fig. 7A). The process is twisted halfway along its marginrunning dorsoventrally and slightly obliquely, and an greateraxis. CAERORHACHIS BAIRDI 239

Figure 6 Caerorhachisbairdi Holmes& Carroll: (a)MCZ 2271,left mandibular ramus, lateral aspect; (b) MCZ 2271,left mandibular ramus, mesial aspect; (c) MCZ 2271,close-up view of parasymphysial region of leftman- dibularramus; scale bars 1 cm. ˆ 240 MARCELLORUTA ET AL.

Figure 7 Caerorhachisbairdi Holmes& Carroll: (a)revised reconstruction, and (b) Holmes & Carroll’s reconstructionof left mandibular ramus, mesial aspect; (c) camera lucida drawing of parasymphysial region of leftmandibular ramus; scale bars=1 cm. CAERORHACHIS BAIRDI 241

Figure 8 Reconstructionsof right mandibular rami of early tetrapods, mesial aspect and to the same length; Meckelianforamina have been blackened; denticle ® elds,when present, have been omitted: (a) Acanthostega gunnari; (b) Crassigyrinusscoticus ; (c) Megalocephaluspachycephalus ; (d) Balanerpetonwoodi ; (e) Eryops megacephalus ; (f) Benthosuchussushkini ; (g) Kotlassiaprima ; (h) Ventastegacuronica ; (i) Eoherpetonwatsoni ; (j) Caerorhachisbairdi ; (k) Gephyrostegusbohemicus ; (l) Pholiderpetonscutigerum ; (m) Archeriacrassidisca ; (n) Discosauriscusaustriacus .Redrawnand modi® ed from the following sources: a± c, h, k, Ahlberg & Clack 1998;d, Milner &Sequeira1994; e, Romer 1947; f, Schoch& Milner 2000;g, Bystrow 1944;i, Smithson1985b; j, this paper;l, Clack1987b; m, Holmes 1989; n, Klembara1997. 242 MARCELLORUTA ET AL. 3.4.5.Dorsal aspect and mesial extension of dentary. In lateral oftheparasymphysial plate is covered with a scatterof small, aspect(Fig. 6A), the dentary deepens uniformly in an anterior posteriorlydirected, irregularly distributed, and heavily worn direction.In dorsal view, it widens abruptly at a pointlying denticles.The plate is approximately subtrapezoidal, but its almostimmediately anterior to the adductor fossa, and coin- lateralcontact with the dentary is not clearly visible, due to cidingapproximately with the position of the ® fthor sixth theheavy disruption of the right ramus, and to the displace- mostposterior tooth. More anteriorly, the width of thedentary mentof a parasymphysialfang in the left ramus. Mesially, isapproximately constant, but it increases slightly just posterior theplate contacts a shortanterodorsal portion of the mesial tothe parasymphysial plate (see below), at the level of the laminaof the splenial, whereas posteriorly, it is probably eighthor ninth most anterior tooth. The dorsolateral margin suturedwith the anterior coronoid. The reconstruction of the ofthe dentary forms a thin,subvertical wall, the mesial side symphysialregion is based on theestimated forward extension ofwhichcovers about two-thirds of theheight of the teeth on ofthe anteriormost part of the coronoid, especially visible in theposterior ® fthof thedentary; and about half of theheight theright ramus. The corresponding region of the left ramus is ofthe remaining dentary teeth. The vertical wall decreases toofragmented to allow us to draw accurate comparisons abruptlyin height near the jaw symphysis. A fewtooth sockets, withits counterpart. The ventral contribution of the anterior visiblealong the length of thedentary, appear as labiolingually endof the dentary to the symphysial region cannot be elongatedepressions delimited by a poorlypronounced, ascertained. peripheralrim. Below therim, the mesial surface of the dentary 3.4.7.Contributions of infradentariesto Meckelian foramina. isirregular, especially near the bases of the largest tooth Inthe anterior third of the jaw ramus, the mesial lamina of crowns.Here, short and low thickeningsthat are obliterated thesplenial contributes to a largearea of the mesial surface rapidlyin a mesialdirection appear close to the sutures immediatelybehind the symphysial region (Figs 4A, 6B, C, formedby the dentary with the coronoid series. The recon- 7A).The dorsal half of the mesial lamina of the splenial is structionof themesial extension of thedentary is problematic. extensivelydenticulated and is delimited from its smooth, Disruptionis especially evident in the right ramus, where lowerhalf by a neatline. The mesial lamina of the splenial thecoronoids are folded mesially along their dorsal margins contributesto the anterior and dorsal margin of the anterior (Fig. 4A). Meckelianforamen. Three small, subelliptical foramina are Thedentary teeth (Figs 6B, 7A) are rather weak and rela- visibleimmediately anterior to the anterior Meckelian foramen. tivelysmall compared with the size of the lower jaw. About Theyoccupy a shallow,anteroposteriorly elongate region just 62tooth positions are observed, with two position peaks, at belowthe denticulated area of the mesial lamina of thesplenial, thelevel of the 22nd and 36th teeth approximately. The andmay correspond to similarly positioned, small foramina in numberof dentary teeth compares well withvalues recorded Megalocephalus,Crassigyrinus , and Gephyrostegus (Ahlberg & insuch Devonian taxa as Acanthostega and Ventastega Clack1998) (Fig. 8). The irregular suture between the mesial (Ahlberg& Clack1998), but is slightly larger than those laminaeof the splenial and postsplenial is visible below the observed in Pholiderpeton (Clack1987b) and Archeria (Holmes anteriorhalf of the ventral margin of the anterior Meckelian 1989),and exceeds greatly those of Greererpeton (Smithson foramen.The suture runs initially very close to the ventral 1982), Crassigyrinus (Panchen1973, 1985), Megalocephalus marginof the foramen before turning ventralward more pos- (Beaumont1977), Proterogyrinus (Holmes1984), Gephyro- teriorly.The mesial lamina of the postsplenial contributes to stegus (Carroll1970), Ichthyostega (Jarvik1980, 1996), early theventral margin of the mid Meckelian foramen and to temnospondyls(Milner & Sequeira1994), all microsaurs, mostof the ventral margin of the posterior Meckelian foramen. nectrideansand seymouriamorphs (Carroll & Gaskill1978; Atthelevel of theposterior Meckelian foramen, a shortoblique Klembara1997; Bossy & Milner1998). The teeth of Caeror- linemarks the posterior boundary of themesial lamina of the hachis areconical, slightly curved posteriorly and lingually, postsplenial.The suture between the mesial lamina of thepost- andtaper smoothly near their blunt tips. Deep striations run splenialand the prearticular is not clearly visible. A thinline alongmost of the crown height, but are usually very faint or runningbetween the anterior and mid Meckelian foramina is absenton its upper third. The crowns are circular in cross- tentativelyinterpreted as partof sucha suture.Likewise, a dis- section,except near their bases, which are slightly elongate ruptedline coming o V theposterior margin of the posterior labiolingually. Meckelianforamen, and continuing irregularly posterodor- 3.4.6.Symphysial region. Thenew reconstruction of the sally,may represent the boundary between the mesial lamina symphysialregion (Fig. 7A) di Versfrom Holmes & Carroll’s ofthe angular and the prearticular. It becomes indistinguish- (1977)(Fig. 7B) in showing a parasymphysialplate inserted ablefrom the network of fractures visible behind the adductor betweenthe anteriormost part of the mesial lamina of the fossa.It is, therefore, impossible to ascertain whether the dorsal splenial(mesially), the anteriormost part of thedentary (later- partof theangular was wrapped around the posterior margin ally)and the anterior coronoid (posteriorly) (Figs 4A, 6B, ofthe jaw ramus and exposed mesially. The area of the C,7C).The hypothesised pattern of fractures on theleft sym- mesiallamina of the angular lying ventral to the posterior physialregion of thejaw ramus matches almost perfectly that Meckelianforamen is occupied by a curved,elongate groove ofthecorresponding area on the right jaw ramus, thus casting seeminglyending posteriorly in a foramen,and by two smaller doubton theaccidental nature of these ‘fractures’.We believe ellipticalforamina. Similar foramina are also observed in other themto represent parasymphysial plates. The parasymphysial taxa(e.g. Clack 1987b; Ahlberg & Clack1998). platecarries two massive, slightly curved fangs lying mesial to 3.4.8.Coronoid series and prearticular. Asin several other theenlarged, anteriormost dentary teeth. In the left ramus, Palaeozoictetrapods, most of the mesial surface of the lower onlyone fang is visible, broken o V one-quarterof the length jawis dominated by the coronoid series and the prearticular, ofthecrown above its base. The remaining part of the crown andonly to a limitedextent by the infradentaries (Figs 4A, liesin close proximity, and is tilted backward, thus obscuring 6B,C, 7A).The mutual relationships of thesebones resemble ashallowsubcircular cavity lying behind the fang insertion. themorphological condition of Eoherpeton (Fig.8). The two Sucha cavityis likely to represent a replacementpit for a mostremarkable features of the coronoids are the large size secondfang. In the right ramus, a wornfang crown and a andgreat elongation of the anterior coronoid, and the fact moreposterior, broken crown are visible. The whole surface thatthe posterior coronoid is relatively short and sends a CAERORHACHIS BAIRDI 243 longanterodorsal process wedged between the dentary and the 3.5.Braincase middlecoronoid. However, it is impossible to establish the pre- Theparasphenoid (Figs 3B, 5A, B) resemblesthose of Eucritta, ciseoutline of this process and its forward extension. Here, it is Crassigyrinus ,Whatcheeria , ‘anthracosaurs’,and someamniotes speculativelyreconstructed as terminating a shortdistance s.s.inshowing a distinctcentral depression (Coates 1996; Clack behindthe posterodorsal angle of the anterior coronoid, 1998a,b2001; Paton et al.1999).It di Versfrom the ¯ atand, insteadof extending forward to contact the latter, as in often,transversely expanded plate found in several temno- Holmes& Carroll’s(1977) reconstruction (Fig. 7B). All three spondyls.It is di cult to ascertainthe presence of pronounced coronoidslack teeth and fangs, but are extensively covered in posterolateralbasal tubera, such as thoseobserved in ‘anthra- tiny,backward-pointing, striated denticles. As in the case of cosaurs’(e.g. Clack 1987b; Holmes 1984, 1989), although thepalatal denticles, striations occur regardless of the size theseseem to have been absent. Only the most proximal part andposition of the denticles, although surface details of the ofthecultriform process is preserved. As aresult,its forward crownare often obscured by poor preservation or breakage. extensionand relationships with the vomers cannot be recon- Theprearticular is more than half as long as the jaw ramus. structed.Possible basioccipital remains are represented by a Itsmesial surface is divided neatly into a denticle-covered rectangularfragment of bone covered in scattered denticles, dorsalarea and a smoothventral area. The boundary between visiblebehind the basal plate, and seemingly articulated with thesetwo areas is marked by anelongate, low thickeningwith thelatter. Although the basal plate is only partially preserved, irregularmargins, running from the posterodorsal margin of itslateral margins and ventral surface do notshow any traces of theanterior Meckelian foramen to a pointsituated just foraminafor the internal carotid arteries. Yates & Warren behindthe sharp bend in the ventral margin of the adductor (2000)discussed four morphological conditions describing the fossa.It then disappears almost abruptly in a posterodorsal positionof the arteries with respect to the basal plate of the direction.Irregular patches of less densely spaced, smaller parasphenoidin various ‘higher’temnospondyls. The absence denticlesare visible along most of the length of the posterior offoramina in Caerorhachis suggeststhat the internal carotid marginof the adductor fossa. The wedge-like anterior end of arteriesdid not enter the parasphenoid through its ventral sur- theprearticular contacts the anterior and mid coronoids, face(foramina are present as a primitivecondition in various dorsally,and the anterior coronoid and mesial lamina of the ‘higher’temnospondyls). As insome temnospondyls, the inter- splenial,anteriorly. nalcarotid arteries probably entered the parasphenoid from the Theposterior extension of the denticles on themesial surface dorsalsurface without passing through ventral foramina on ofthe lower jaw of Caerorhachis isa ratherunusual feature. theplate. However, an alternative possibility is suggested by Acomparablerearward extension has been recorded in very theoccurrence of an anteroposteriorlyelongate gap in the left fewtetrapods. Beaumont (1977, p. 71) described and recon- halfof the denticle-covered area of the basal plate (Figs 3B, structed ‘...anextensionof thecoronoid denticles forming a 5A,C), visibleimmediately behind the central circular depres- narrowband’ along the posterior two-thirds of the dorsal sion.Clack (2001) interprets a similarlypositioned gap in the marginof the prearticular in Megalocephalus . Such a band basalplate of Eucritta asthe pathway of theleft internal carotid wouldextend posterodorsally to a pointimmediately below the artery.The condition of Caerorhachis compareswell withthat articular.In contrast, Ahlberg & Clack(1998, p. 33) ‘. . . draw of various ‘higher’temnospondyls, except for the more pos- attentionto the absence of shagreen on the coronoids or teriorposition of the carotid pathway (Schoch & Milner prearticular,and its replacement by ® nebut irregular pit- 2000;Yates & Warren2000). tingand sculpturing’ on the mesial mandibular surface of Megalocephalus . Loxommarankini likewiseshows a scatterof 3.6Axial skeleton posterodorsaldenticles on the prearticular, not discussed by Thevertebrae conform to, and are in several respects simpler Beaumont(1977), but mentioned by Ahlberg & Clack(1998). (?moreprimitive) versions of, the gastrocentrous pattern Outsidetetrapods, an extensively denticulated prearticular exhibitedby ‘anthracosaurs’.The ossi® ed portions of inter- occursin various actinopterygians (e.g. Gardiner 1984) and centraand pleurocentra (Figs 9, 10A,B) resemblemore lightly sarcopterygians(e.g. Jarvik 1980; Fox et al. 1995; Long et al. builtversions of thoseof Gephyrostegus (Carroll1970), both in 1997). generalproportions and in surface details, including sculptur- 3.4.9.Articular. Thearticular (Figs 4A, 6B, 7A), about twice inga extensionof thearticular surfaces. They are most similar aswideas long,sits on topof theposterodorsal margin of the tothose of the early ‘anthracosaurs’ Silvanerpeton (Clack1994) surangular,its mesial edge being slightly anterior and ventral to and Eldeceeon (Smithson1994). Similarities with these taxa are itslateral edge. In dorsal aspect, its outline is approximately alsoevident in the proportions of the trunk neural spines. These bean-shaped,with thick, blunt-topped, anterior concave aresquare and stout, often only slightly taller than long in their marginand posterior convex margin. Its mesial margin is upperhalf (above the zygapophyses level), and unlikethe com- bluntlyconvex and evenly round in dorsal view, whereas its parativelymuch taller spines of Proterogyrinus and Archeria lateralmargin delimits a stout,roughly triangular process. The (Holmes1984, 1989). The irregular, almost jagged upper anteriorand posterior margins delimit a trough-likecondyle, marginof some spines indicates that these were covered in a stronglyconcave parasagittally, and gently convex trans- cartilagecap in life. The cervical ribs di Verremarkably from versely.In lateral aspect, there appears to be nodistinctbound- thetrunk and caudal ribs (Fig. 10C± E). Theyare straight, arybetween surangular and articular. In posteriorand mesial andcarry ¯ atteneddistal ends, a patternobserved, among views,a narrowgroove is discernible below the ventral surface others,in several basal and derived ‘anthracosaurs’(Holmes ofthe articular. Posterodorsal to it, the surangular sends a 1984,1989; Clack 1994; Smithson 1994; Smithson et al. small,subtriangular lappet around the bowl-shaped base of 1994),some microsaurs (Carroll & Gaskill1978), and seymour- thearticular. The suture between the articular and the pre- iamorphs(White 1939; Berman et al.2000).This morpho- articularis not clearly identi® able. In mesial aspect, the logicaldistinction was used by Coates (1996) to characterise posterodorsalregion of the jaw ramus lying immediately asubsetof the ‘reptiliomorph’branch of his cladogram. The belowthe articular probably represents the rearmost and midtrunk ribs (Fig. 10D) are unlike those of ‘anthracosaurs’ dorsalmostpart of the prearticular. It is di cult to ascertain (includingsuch early forms as Eldeceeon and Silvanerpeton ), towhatextent the articular contributed to this region as well. inbeing comparatively much shorter, and exhibiting a more Achordatympani foramen could not be seen. abruptreduction in absolute size in an anteroposterior 244 MARCELLORUTA ET AL.

Figure 9 Caerorhachisbairdi Holmes& Carroll, axialand appendicular skeleton of MCZ 2271;scale bar 5 cm. ˆ CAERORHACHIS BAIRDI 245

Figure 10 Caerorhachisbairdi Holmes& Carroll: (a)pleurocentrum, anterior aspect; (b) pleurocentrum, poster- ior aspect;(c) left cervical rib, dorsalaspect; (d) left mid trunk rib, posterodorsalaspect; (e) left posterior trunk rib, dorsalaspect; (a± e) tothe same scale; (f) neuralspine outlines, left aspect and to the same scale; (g) haemal spine outlines,left aspect and to the same scale. All camera lucida drawings. Scale bars=l cm for (a±e); 0 :5cmfor (f, g). 246 MARCELLORUTA ET AL. direction.Marked changes in neural spine morphology, as well ter.A thinsheet of bone extends from the dorsal iliac blade to asin the relative o Vsettingof pre- and postzygapophyses, are therobust, straight posterodorsal process. Compaction makes seenat three levels (Figs 1A, 9, 10F): between the anterior itimpossible to ascertain to which degree did the posterodorsal andposterior halves of the presacral column (corresponding processlie lateral to the dorsal blade in life. Likewise, it is topresacral 15 or 16, approximately); between the posterior- dicult to establish whether the dorsal process was un® nished mostpresacrals and the most anterior caudals; and between anteriorly.A distincttransverse ridge (Romer 1922), inter- themost anterior caudals and the preserved portion of tail pretedby Coates (1996) as a possibleapomorphy for basal carryinghaemal spines. Drastic height and length reduction amniotes,sweeps backward along a straightcourse from the ofthe most posterior caudal neural spines is also observed lateralto the ventral surface of theposterodorsal process. We (correspondingto caudal 14 or 15, approximately). The pointout that a similariliac ridge was also ® guredby Romer haemalspines (Figs 9, 10G)are comparable in relative size to (1922) in Eryops (see ‘Remarks’below), and by Holmes et al. thoseof the vast majority of ‘anthracosaurs’,except such (1998) in Dendrerpeton .Anelongate, depressed area lies earlyforms as Eldeceeon.Insome of the better-preserved immediatelydorsal to the ridge. Muscle insertions are visible haemalspines, the ventral ends are foreturned and slightly inthe form of irregularly sigmoid striations and crests across expandedanteroventrally. The anterior margin of such an thedepressed area and below the ridge. The iliac neck is expandedend in one particularly well-preserved haemal spine ratherrobust compared with those of ‘anthracosaurs’,and is carriesminute, round bosses. However, preservation makes it oVsetwell posteriorlywith respect to the supracetabular dicult to ascertain whether these bosses are a genuinefeature. buttress(discussion in Coates 1996). Theybear little resemblance to the heavily serrated anterior 3.7.2.Hind-limb. Theleft femur (Fig. 12A) is preserved edgeof the second haemal spine of Acanthostega (Coates mainlyin dorsal (extensor) aspect. Its shaft, head and adductor 1996).The anteriormost haemal spine seems to have a ¯attened, bladeare slightly crushed, but its condylar region is almost subrectangulardistal end. Likewise, the more posterior haemal intact,albeit ¯ attened.Proximal and distal ends are only spinesterminate in aslightlyspatulate end, but areless strongly slightlywider than the shaft, which confers a rathermassive foreturnedthan the more anterior spines. The rearmost pre- aspectto the bone (see Fig. 12B± H, J±Y forcomparisons servedspines show no traces of a spatulateend. withselected femora in other extinct tetrapods). Little torsion isevident around the longitudinal axis. The dorsal surface of thehead is smoothly convex, more so than the shaft and the 3.7.Appendicular skeleton proximalpart of the condylar region. The adductor blade pro- 3.7.1.Pelvic girdle. Afewobservations complement Holmes jectsanteriorly and slightly ventrally from the anterior surface, &Carroll’s(1977) description (Fig. 11). Sutures between ilium, extendingfrom a pointsituated slightly distal to themid length ischiumand pubisare visible inside the acetabulum in the form oftheshaft, to apointsituated immediately distal to thetrans- ofirregular grooves crossing the coarsely granular endo- verselevel of the femoral head. The dorsal surface of the blade chondralsurface. They cannot be tracedon thelateral surface isshallow near its insertion onto the shaft, but becomes ¯ atter ofthe girdle, except for a shorttract of the ilium/ ischium awayfrom the latter, due to compaction.The distal half of the suture,visible posterior and slightly ventral to a short, blademargin shows irregular denticulations, possibly resulting tongue-likenotch in the posterior part of the acetabular rim fromdisruption. The proximal apex of thebladeends in a stout (seeCoates 1996). Behind this notch, the iliac surface shows a boss-likeinternal trochanter, separated from the proximal end deep,subcircular pit, below which is a slightlyraised post- ofthefemur by adeeplyincised trough, a conditionobserved acetabularbuttress. Two small,round bosses and an irregular, in e.g. Tulerpeton (Lebedev& Coates1995), Crassigyrinus elongatethickening are visible inside the pit. The signi® cance of (Panchen& Smithson1990) and Archeria (Romer1957; thesestructures is obscure; they may represent accidental Holmes1989). A small,elongate tuberosity on the postero- bumpswhich occurred during the casting. The supracetabular dorsalsurface of the femoral head, lying approximately at the buttressis robust, and carries a thicklateral margin. Antero- sametransverse level as theinternal trochanter, may represent ventrally,it merges into a poorlypronounced, anterior theischiotrochantericus muscle insertion, based on acompari- tongue-likenotch. The ventral acetabular rim is produced sonwith a similarlypositioned structure in the femur of the intoa semicircularlappet anteriorly, before merging into the synapsid Dimetrodon (Romer1956). The anterodorsal region un®nished, anterior margin of the pubis. The lappet, and the ofthetibial condyle is coveredby theproximal end of the left areaof pubis lying just anteroventral to the acetabulum, are tibia,but it is possible to discernpart of the seemingly bulbous coveredin short, irregular striations (?insertions for ambiens tibialfacet. Similarly, only the dorsalmost part of the ® bular and/orpubotibialis muscles; see Romer 1922, 1956). The stria- facetis visible. The visible part of the trough-shaped, distal tionsin thecentral area of thepubis are weaker than those on surfaceof the intercondylar groove seems to have been made thelappet. A few,faint striations are also visible on theischium, of® nishedbone, but its entire ventralward extension cannot behindthe acetabulum (?insertions for ischiotrochantericus bereconstructed. The tibial condyle is slightly shorter than muscle;Romer 1956). The proportions of the puboischium the® bularcondyle, and shows a morestrongly convex dorsal resembleclosely those of Proterogyrinus and Archeria (Romer surface,as is the case in most Palaeozoic tetrapods (Holmes 1957;Holmes 1984, 1989). However, the relative size of the 1984).The intercondylar groove is fairly distinct, deepens dis- anteriorand posterior tongue-like projections are more similar tally,and carries weak, irregular striations and rugosities, to those of Whatcheeria (Lombard& Bolt1995). A narrowarea especiallyevident on the posterior surface of thetibial condyle. ofendochondralbone is visible anterior to themain acetabular Two orthree crescent-like ridges run transversely across the surface.It isdelimited posteriorly by asmallconstriction of the deeperpart of the intercondylar groove. Small, irregular acetabulumat the level of the ventral lappet and anterior ridgesalso straddle the boundary between the tibial/ ®bular tongue-likenotch. No foramina are visible on the puboischium. facetsand their respective condyles, and are better developed However,details of its lateral surface are di cult to detect onthe distal part of the dorsal surface of the ® bularcondyle. becauseof the dense network of fractures. Thestout left tibia (Fig. 13A, B) isless than half as longas Aremarkablefeature of theilium is the presence of a small, thefemur (40% length ratio; Holmes & Carroll1977). A similar dorsaliliac blade (not discussed by Holmes & Carroll1977), ratiocharacterises such tetrapods as Tulerpeton (Lebedey & regardedby Coates(1996) as apossible ‘reptiliomorph’charac- Coates1995), Crassigyrinus (Panchen& Smithson1990), and CAERORHACHIS BAIRDI 247

Figure 11 Caerorhachisbairdi Holmes& Carroll: (a)MCZ 2271,left femur, left half of pelvic girdle and proximal caudalvertebrae; (b) camera lucida drawing of lefthalf of pelvic girdle; (c) labelled reconstruction of lefthalf of pelvicgirdle; scale bars 1 cm. ˆ 248 MARCELLORUTA ET AL. CAERORHACHIS BAIRDI 249 Watcheeria (Lombard& Bolt1995), among others. It is waspresumably overlapped by the adjacent posterior gastra- lower,however, than the ratio found in Eucritta (Clack lium.Poorly preserved gastralia in dorsal aspect show an asym- 1998b,2001 and ‘anthracosaurs’(e.g. Silvanerpeton,Eldeceeon, metricalarrangement of concentric growth striations. Faint Gephyrostegus,Proterogyrinus and Archeria;Romer1957; striationsare sometimes visible on the spoon-shaped extremity Carroll1970; Holmes 1984, 1989; Clack 1994; Smithson ofgastraliaexposed in ventral view. 1994).The proximal end of the tibia is comparable in size to thecondylar region of the femur. The more deeply concave ofthe two margins of the shaft is likely to represent the 4.Reconstructionand life-style mesialmargin, based on a comparisonwith other early tetra- pods,and indicates the presence of a distinctinterepipodial Therelative size and proportions of the hind-limb, a fully space.The orientation of the tibiale and intermedium facets ossi®ed tarsus, the lack of lateral line canals and lithological onthedistal end of thebone suggests that the tibia is exposed evidencepoint to Caerorhachis asa terrestrialvertebrate, in¯ exorview. A ridgeis visiblein the middle of theproximal althoughthe depth of its tail suggests partially aquatic life halfof the¯ exorsurface (see also Romer 1957; Holmes 1984, habits(Holmes & Carroll1977; Smithson 1980; Carroll 1989;Godfrey 1989). In general proportions and relative thick- 1996).A semiterrestriallife-style is also indicated by the ness,the ridge is similar to a similarlypositioned structure in femur/puboischiumlength ratio (about 109 :13%),which is Eoherpeton (Smithson1985b), and resembles a lessrobust amongthe highest recorded ratios for Palaeozoic tetrapods. andless oblique version of the rugose ridge for ¯ exormuscle Suchratio compares well withthose of suchtaxa as the ‘anthra- attachmentdescribed by Romer (1957) in Archeria, and by cosaur’ Gephyrostegus (106:52%;Carroll 1970), the microsaurs Holmes(1984) in Proterogyrinus . However, Caerorhachis Microbrachis and Tuditanus (105:56% and l22:95%,respec- showsonly a faintindication of the rugosities noted by tively;Carroll & Gaskill1978), the temnospondyl Dendrerpeton Romer(1957) and Holmes (1984). These appear as poorly pre- (105:71%; Holmes et al.1998),and various dissorophoid tem- servedstriations in several areas of the proximal half of the nospondyls,among others. The extensive covering of small ¯exorsurface. There is no posterior tibial ¯ ange,in contrast denticleson the palate and lower jaw of Caerorhachis , and with Crassigyrinus (Panchen& Smithson1990), Tulerpeton thesmall size of dentary and maxillary teeth, are consistent (Lebedev& Coates1995) and Westlothiana (Smithson et al. withswift and/ orslimyprey as possiblefood sources (e.g. inver- 1994). tebratesand/ orsmall vertebrates). MCZ 2271probably Littlecan be added to thedescription of the® bulaand foot reacheda subadultor young adult stage when it died, based (Figs13C, 14). A tarsalelement, lying close to the epipodia, onwell-developed dermal sculpturing, tightly closed skull resemblesa combinedtibiale+centrale 4 asdescribed by sutures,and ossi® ed tarsus (Milner & Sequeira(1994) and Holmes(1984) in Proterogyrinus . The ‘8-shaped’aspect of Berman et al.(2000)discuss similar features in other tetrapods). thiselement, with a distincttrough-like ‘waist’between a smal- Comparisonswith broad-snouted, ¯ at-headed,and short- lersubpolygonal part and a largersubelliptical part, is tenta- ribbedtetrapods (e.g. Balanerpeton and Eucritta) suggest tivelyinterpreted as representing an earlierstage of the fusion that,like these, Caerorhachis probablyused buccal pumping betweentibiale and centrale 4 than that of Proterogyrinus . The forair intake (Milner & Sequeira1994; Clack 2001). Although intermediumcarries a notchwhich gives it a vaguelyL- theskull depth cannot be reconstructedaccurately, short trunk shapedaspect. An L-shaped intermedium is regarded as a ribsand broad palate are consistent with this functional inter- ‘reptiliomorph’character by Lebedev & Coates(1995) and pretation.Long and curved ribs, such as those of ‘anthraco- Coates(1996). saurs’,seymouriamorphs, diadectomorphs and crown-group 3.7.3.Body armour. Severalisolated gastralia are in ventral amniotes(Romer 1956; Carroll 1970; Holmes 1984, 1989; (external)aspect (Fig. 15). Their orientation is based on com- Smithson1985b, 1994; Clack 1994; Smithson et al. 1994; parisonswith gastralia of ‘anthracosaurs’(Panchen 1985; Berman et al.2000;Klembara & Bartik2000), suggest that Clack1987b). A thickridge, decreasing smoothly in height thesetaxa adopted costal ventilation for breathing. Also, they anteromesiallyand posterolaterally, lies close to one of two possessmuch deeper rib cages than Caerorhachis , as well as longmargins of each gastralium. In several gastralia that moreelongate and deeper skulls. Size and curvature of the occupythe ventral midline and the ¯ anksof the animal, one trunkribs in Caerorhachis probablymade them unsuitable extremitytapers gently and is approximately triangular, forexclusively costal breathing. The curved ribs of many whereasthe other is abruptly truncated and broadly spoon- stereospondylsare likely to be a size-relatedfeature (Milner shaped.Similar di Verenceshave been recorded in the gastralia 1988,1990) in taxa that are otherwise adapted for buccal ofother tetrapods, e.g. Crassigyrinus (Panchen1985). Some pumping. gastraliaappear subrectangular. Posterior to the ridge, the Head/trunkproportions in Caerorhachis comparewell with ventralsurface of the gastralium is shallowly depressed, and those of Ichthyostega (Jarvik1980, 1996; Carroll 1988), Eucritta

Figure 12 Caerorhachisbairdi Holmes& Carroll: (a)camera lucida drawing of left femur, extensor view; (b± y) reconstructionsof leftfemora of earlytetrapods, extensor view and to the same length: (b) Acanthostegagunnari ; (c) Ichthyostegastensioi ; (d) Tulerpetoncurtum ; (e) Greererpetonburkemorani ; (f) Whatcheeriadeltae ; (g) Crassigyr- inusscoticus ; (h) Eoherpetonwatsoni ; (i) Caerorhachisbairdi ; (j) Archeriacrassidisca ; (k) Proterogyrinusscheelei ; (l) Westlothianalizziae ; (m) Dimetrodonlimbatus ; (n) Discosauriscusaustriacus ; (o) Seymouriabaylorensis ; (p) Limnoscelisdynatis ; (q) Eldeceeonrolfei ; (r) Rhynchonkosstovalli ; (s) Trachystegosmegalodon ; (t) Tuditanus punctulatus ; (u) Sauropleurascalaris ; (v) Balanerpetonwoodi ; (w) Mastodonsaurusgiganteus ; (x) Acheloma sp.; (y) Scincosauruscrassus .Redrawnand modi® ed from the following sources: b, Coates 1996; c, Jarvik 1996; d,Lebedev& Coates1995; e, Godfrey 1989; f, Lombard& Bolt1995; g, Panchen & Smithson1990; h, Smithson 1985b;i, this paper;j, Romer1957; k, Holmes 1984; 1, Smithson et al.1994;m, o,Romer 1956; n, Klembara & Bartik2000; p, Berman & Sumida1990; q, Smithson 1994; r± t, Carroll &Gaskill1978; u, y, Bossy & Milner 1998;v, Milner &Sequeira1994; w, Schoch& Milner 2000;x, Sullivan et al.2000.Scale bar 1 cm. ˆ 250 MARCELLORUTA ET AL.

Figure 13 Caerorhachisbairdi Holmes& Carroll: (a)MCZ 2271,left tibia, ® bula,and foot; (b) camera lucida drawingof lefttibia and ® bulain ¯ exorview; (c) labelled diagram of left tibia and ® bula;scale bars 1 cm. ˆ CAERORHACHIS BAIRDI 251

Figure 14 Caerorhachisbairdi Holmes& Carroll: cameralucida drawing of leftfoot; scale bar 1 cm. ˆ

(Clack1998b, 2001) several temnospondyls (e.g. Cacops, identifyingthree cranial synapomorphies for the group: (1) pre- Eryops,Dendrerpeton ,manystereospondyls, etc.; Williston maxillawith marginal anteroposterior elongation and broad 1910;Romer 1947, 1966; Moulton 1974; Holmes et al. 1998), dorsomedialsurfaces; (2) jugal/ prefrontalsuture excluding some ‘anthracosaurs’(e.g. Gephyrostegus ;Carroll1970), the lacrimalfrom orbit margin; (3) substantial jugal contribution microsaur Pantylus (Carroll& Gaskill1978), certain diplo- topreorbital region through development of broad jugal/ caulidnectrideans (A. C. Milner1980; Bossy & Milner1998), lacrimalsuture. Unfortunately, Caerorhachis cannot be etc.Interestingly, the presacral vertebral count of these taxa assessedfor any of these characters. rangesbetween 22 and 24, except in Caerorhachis (32), and in Despitehis reservations about the temnospondyl nature of diplocaulids(fewer than 18). Caerorhachis ,A.R.Milner(1980) did not rule out the pos- sibilitythat it might represent the juvenile of either a large, long-snouted,and Edops-likeform, or a moreprimitive 5. A nities ‘pre-edopoid’temnospondyl. This interpretation was based on suchmorphological similarities between Caerorhachis and 5.1.Holmes &Carroll’sinterpretation Edops as ‘caninepeaks’, mediodorsal orientation of articular, Beforediscussing the inclusion of Caerorhachis in various andshape of pterygoids and palatal vacuities. However, cladisticanalyses, we summariseprevious researchers’ views ‘caninepeaks’ are observed in other tetrapods (e.g. Crassi- onthea nities of this tetrapod. gyrinus, Megalocephalus , Pholiderpeton , Proterogyrinus , etc.; 5.1.1. Is Caerorhachis an edopoid? Holmes& Carroll Panchen1973, 1985; Beaumont 1977; Holmes 1984; Clack (1977)assigned Caerorhachis ,withsome uncertainty, to the 1987b,1998a). In addition, both jaw rami of Caerorhachis familyDendrerpetontidae, superfamily Edopoidea. However, haveundergone disruption and ¯ attening,causing the articular A.R.Milner(1980) viewed characters used by them to ally toface almost entirely mesially. dendrerpetontidsto edopoids as plesiomorphic for temno- Thesimilar morphology of pterygoids and palatal vacuities spondylsor even for tetrapods. His emended Edopoidea in Caerorhachis and Edops deservessome comment. Knowl- excluded Dendrerpeton ,regardedas one of the most primitive edge of the Edops palaterests mostly on a singlelarge andgeneralised temnospondyls (Romer 1947; Berman et al. skull(Romer & Witter1942). Its spindle-like palatal vacuities 1985;Godfrey et al.1987;Milner & Sequeira1994; Yates & resemblenarrower, and more elongate versions of those of Warren2000; but see Holmes et al.1998).Edopoids are Caerorhachis . In Edops,andperhaps also in Caerorhachis , the medium-to large-sized, early-mid Pennsylvanian to earliest pterygoidsmeet anteriorly along a mediansuture (see ‘Descrip- Permiantemnospondyls resembling alligators or gharials in tion’above). This feature may be related to the large size and skullmorphology (Milner 1990; Milner & Sequeira1998). broadpreorbital region of Edops (A.R.Milner1980). It may Recentreviews (Sequeira & Milner1993; Milner & Sequeira havereinforced the anterior third of thepalate, counteracting 1994,1998; Sequeira 1996) clarify their limits and content, bitingforces generated when the animal fed upon large prey, 252 MARCELLORUTA ET AL. andextension of the quadrate and palatine rami of the ptery- goids,and con® guration of theparasphenoid. 5.1.2. Is Caerothachis atemnospondyl? Apartfrom palatal vacuities,hardly any character of Caerorhachis suggests resemblanceto Edops orother temnospondyls. According to A.R.Milner(1980, 1990), if Caerorhachis isplaced within temnospondyls,then several cranial and postcranial features ofthis tetrapod appear to be autapomorphic. As such, they cannotestablish its relationships to other temnospondyls. Furthermore,some of the characters employed by Holmes & Carroll(1977) to support the temnospondyl a nities of Caeror- hachis areeither primitive for tetrapods (e.g. dermal sculpture; supratemporal/postparietalsuture; Milner & Sequeira1994; Coates1996), or ofuncertainpolarity (e.g. solid fusion between cheekand skull table; uniramous ilium; Milner 1993b; but see Coates1996). Smithson (1982) placed emphasis on the occur- renceof enlargedpleurocentra to questionHolmes & Carroll’s (1977)interpretation. Holmes (1984) acknowledged that the vertebraeof various ‘anthracosaurs’show little resemblance tothe cylindrical centra of amniotes s.s.,microsaursor the temnospondyl Doleserpeton .Hesuggested that the morpho- logicalcondition of Caerorhachis ,althoughsimilar to that of certain ‘anthracosaurs’,probably represents an early stage in thedevelopment of gastrocentrousvertebrae in temnospondyls. Doleserpeton woulddisplay the culmination of such a develop- ment.The distribution of gastrocentrous vertebrae in Palaeo- zoictetrapods would imply that enlarged pleurocentra developedconvergently in temnospondyls and ‘anthracosaurs’, although ‘ ...[the]speci® c relation[of Caerorhachis ], if any, to Doleserpeton isunclear’ (Holmes 1984, p. 507). However, Doleserpeton isusually regarded as one of the most derived temnospondyls,and is placedby someauthors as theimmedi- atesister taxon of crown-group lissamphibians (Bolt 1969, Figure 15 Caerorhachisbairdi Holmes& Carroll: reconstructed 1991;Milner 1988, 1990). Therefore, its close relationship gastraliumin ventral aspect; scale bar 1 mm. ˆ with Caerorhachis isimplausible. Recently,Yates & Warren(2000) presented a cladistic analysis of ‘higher’temnospondyls (i.e. exclusive of edopoids, Dendrerpeton and Balanerpeton ).Ofthe two apomorphies assuggested by analysis of stress distribution in the skull of usedby them to assess their ingroup monophyly ( ®de Laurin moderncrocodiles (Busbey 1995). Such a strengtheningfunc- &Reisz1997, and Laurin 1998a; but see Bolt 1979, 1991, and tion,however, is di cult to reconcile with the much smaller Milner1988, 1990, 1993b), the ® rst(loss of entepicondylar fora- size of the Caerorhachis skull. men)cannot be evaluated in Caerorhachis ,whereasthe second Holmes et al.(1998)noted that, in those temnospondyls in (posteriorcoronoid visible in lateral view) is absent. However, whichgrowth series are known, the palatal vacuities undergo thereremains a possibilitythat Caerorhachis fallsoutside Yates [almost]isometric changes. If this is true for Caerorhachis and &Warren’s(2000) ingroup, and represents a moreprimitive Edops,thenA. R.Milner’s(1980) hypothesis that the former temnospondyl.Milner (1988) used six characters to de®ne the mayrepresent a juvenile,edopoid(-like) temnospondyl cannot basaltemnospondyl node, including taxa not considered by bedismissed altogether. Obviously, lack of additional fossil Yates& Warren(2000): (1) manus with four digits; (2) semi- materialmakes it impossible to test such a hypothesis.How- circulartympanic notch not reaching quadrate, and dorso- ever,elongate and spindle-shaped palatal vacuities are laterallyoriented stapes; (3) round interpterygoid vacuities at presentin several other Palaeozoic taxa, such as colosteids leasthalf as wideas skulland bordered by triradiatepterygoids; (Smithson1982; Hook 1983), various nectrideans (Bossy & (4)broad, ¯ atvomers widely separating choanae; (5) straight Milner1998), many microsaurs (Carroll & Gaskill1978), andshort ribs; (6) parasphenoid with parallel-sided, elongate some ‘anthracosaurs’(e.g. Carroll 1970; Holmes 1984, 1989), cultriformprocess overlapping or suturing with vomers and etc.The validity of this character is, therefore, questionable. withbroad, ¯ atbasal plate. Two ofMilner’s(1988) characters Furthermore,the insertion of the anterior ends of the palatal (1,2) cannot be observed in Caerorhachis ;three(3± 5) are ramiof the pterygoids between the vomers is also documented notmatched by its morphology; and one (6) can be evalu- incertain microsaurs and nectrideans, among others. The atedonly partially, although it does not seem to suggest any recentlydescribed, primitive edopoid Adamanterpetonohioensis resemblanceto temnospondyls (see ‘Description’above). An (Milner& Sequeira1998) may represent, in some respects, a additionalfeature employed by Milner(1990, 1993b) to charac- bettermodel for inferring the juvenile condition of Edops terisetemnospondyls, not considered further by Milner & than Caerorhachis does.Although the palatal con® guration Sequeira(1994), is the occurrence of a posterolateral¯ ange of Adamanterpeton issuper® cially similar to that of Caeror- onthepalatine ramus of thepterygoid, not present in Caeror- hachis (e.g.presence of relatively small, albeit round, inter- hachis.Arudimentary¯ angeon the palatine ramus is also pterygoidvacuities; exclusion of vomers and palatines from observedin some microsaurs and nectrideans (Carroll & suchvacuities), it di Versfrom the latter in the greatlyexpanded Gaskill1978; Bossy & Milner1998), but it is rarely as well vomers,shape and relative size of the choanae, morphology developedas those of various temnospondyls. The ¯ ange CAERORHACHIS BAIRDI 253 confersa characteristicanteromesial constriction to the sub- 5.3.Coates’ 1996 matrix temporalfossa (Romer 1947), and was lost twice during the Coates’(1996) study (18 taxa; 76 characters; 36 characters evolutionaryhistory of temnospondyls, according to Yates & unscorablefor Caerorhachis )elaboratesupon Lebedev & Warren(2000; but see Milner 1990). Other characters discussed Coates’s(1995) work, and traces the lissamphibian/ amniote byMilner(1990) and Milner & Sequeira(1994) in thecontext splitback to Palaeozoic taxa. PAUP* ®ndssix trees at 144 ofbasal temnospondyls relate to premaxillary shape (not steps (C.I. 0:5839; R.I. 0:7077; R.C. 0:4276)that place observed in Caerorhachis )andpresacral vertebral count. The Caerorhachisˆ assister group ˆ of Westlothianaˆ .Optimisationof vertebralcount of Caerorhachis exceeds25, considered by scorablecharacter changes for ( Caerorhachis Westlothiana ) Milner(1990) to be primitivefor temnospondyls. isas follows: 6 (absenceof tooth row on ‡parasymphysial Caerorhachis diVersfrom temnospondyls also in details of plate),7 (absenceof parasymphysial foramina), 58 (absence thelower jaw (Fig. 8). Thus, in several early and many later oftabular horn with super® cial and deep components). temnospondyls,the shallow, anterior two-thirds of the lower Reweightingcharacters yields three trees, but does not a Vect jawramus is delimited more or lesssharply from its posterior therelationships of Caerorhachis . The (Westlothiana Caeror- thirdby a suddendecrease in jaw depth (Milner & Sequeira hachis)cladehas a low bootstrapsupport (60%) and ‡ decay 1994,1998; Holmes et al.1998),regardless of the absolute index(1). The ( Archeria (Proterogyrinus Westlothiana )) sizeof the lower jaw. The ventral margin of the anterior two- branchingsequence is the ‡reverse of thatfound ‡ in the original thirdsmay be straight,gently concave or even sinuous in lateral analysis. aspect,and merges with a broadlyconvex, or sharply bent Inspectionof the published version of Coates’matrix reveals posteroventralangle. The upper dentary margin is often minorinaccuracies related to four characters. The states of straightwhen viewed laterally, except for a short,sloping characters65 and74, describingthe occurrence of apentadactyl portionanterior to the surangular/ coronoidcrest (Ahlberg & anda tetradactylmanus, respectively, are clearly comple- Clack1998). The dorsolateral margin of the adductor fossa is mentaryfor all of Coates’ post-Devonian tetrapods in which subhorizontaland often parallel to the dentary margin. By themanus is known. In the original matrix, Westlothiana is contrast,the lower jaw of Caerorhachis issimilar in general codedas 1 forcharacter 65, and as unknown for character proportionsto thoseof the ‘anthracosaurs’ Archeria and Pho- 74.Obviously, the cell entry for character 74 should be 0, liderpeton (Clack1987b; Holmes 1989), or thebaphetid Mega- since Westlothiana isassumed to havepossessed a pentadactyl lophalus (Beaumont1977). Its pro® le is smooth along its ventral manus(though its digit count is not certain; Smithson, pers. andposterior margins, whereas the dentary margin is gently comm.).There is partial redundancy associated with characters concave.Additional di Verencesbetween Caerorhachis and 21and35.Character 21 refersto theabsence or presence of cor- temnospondylsare observed at thelevel of the Meckelian for- onoidfangs. Character 35 refers to the occurrence (or lack aminaand of the mesial laminae of the infradentaries. Thus, thereof)of the condition ‘coronoidfangs lost’. In the original exceptfor some taxa (e.g. Milner & Sequeira1994), several tem- matrix,the codings related to these two characters are comple- nospondylsshow a small,subelliptical or circular posterior mentaryfor most taxa. However, Whatcheeria iscoded as 0for Meckelianforamen between prearticular, angular and post- character21, andas unknown for character35. Aftercharacter splenial,or between prearticular and angular. Often, a small, 35is removed, the amended version of Coates’ matrix yields anteriorMeckelian foramen is visible between prearticular, 20trees(Fig. 16A; strict consensus). The characters supporting splenialand postsplenial, or betweensplenial and postsplenial. theposition of Caerorhachis areunchanged. Caerorhachis resembles Eoherpeton inshowing a rowof three foraminabetween the mesial laminae of the infradentaries 5.4.Paton, Smithson & Clack’s1999 matrix andthe prearticular. Such foramina are intermediate in relative Paton et al.(1999)analysed a modi®ed and enlargedversion of sizebetween those of Eoherpeton and those of Archeria,Pholi- Clack’s(1998a,b) data sets (78 cranial, 33 postcranial charac- derpeton and Proterogyrinus (Fig.8). As in these taxa, the ters;55 characters unscorable for Caerorhachis ) in order to rearmostextension of the mesial lamina of the splenial in placesystematically Casineria.PAUP*® nds18 trees at 345 Caerorhachis appearsto be closer to the anterior margin of steps(Fig. 16B; strict consensus). Caerorhachis ispaired with theadductor fossa than to the symphysial region of the man- ‘anthracosaurs’in all minimal trees, albeit with low decay dible.This condition is found also in other taxa (see ‘Conclud- index(1) and bootstrap support (17%). As an example of ingremarks’). characteroptimisation, in those trees in which Whatcheeria, Eucritta and Baphetes areintermediate in position between 5.2.Other interpretations ‘anthracosaurs’and Seymouria,theposition of Caerorhachis Milner& Sequeira(1994) noted similarities between Caeror- issupported by the following scorable characters: 8 (lateral hachis and Baphetes inthe con® guration of the anterior part edgeof intertemporalnot interdigitating with cheek), 25 (post- ofthe palate, in the shape and position of the choanae, and orbitalsuture to skulltable (intertemporal) smooth), 77 (skull inthe proportions of the palatines and vomers. They hypo- tablelonger than broad), 98 (presence of V-shaped grooves thesisedthat Caerorhachis representsa gradeof anatomical onendsof phalanges),and 104 (one or twopairs of postsacral organisationbelow temnospondyls comparable to that of ribsapproaching length of trunk ribs). Character reweighting baphetids. yieldsa singletree (Fig. 16C), but does not a Vectthe position Morerecently, Coates (1996) suggested that Caerorhachis of Caerorhachis . Removing Eucritta givestwo trees at 334 maybe more closely related to amniotes than to lissam- steps(Fig. 16D) in which Casineria isforcedout of theamniotes phibians,based on theshape of the centra, the morphological s.s.,andplaced with ‘anthracosaurs’in two di Verentpositions. distinctionbetween cervical and trunk ribs, the greatly reduced Characterssupporting the position of Caerorhachis are the dorsalprocess of the ilium, a transversepelvic ridge sweeping samein both trees, and most of them match the apomorphy towardsthe ventral edge of the posterior iliac process, and an listfound in the original parsimony run. L-shapedtarsal intermedium. Based on inferred character optimisationon his preferred cladogram, Coates (1996) 5.5.Ahlberg & Clack’s1998 matrix suggestedthat Caerorhachis ismore crownward than West- Ahlberg& Clack’s(1998) analysis of lowerjaw characters tried lothiana onthe amniote stem-group, and basal with respect to toassess the relative positions of early taxa known only from ‘anthracosaurs’. theirmandibles, and discussed key characters that may allow 254 MARCELLORUTA ET AL.

Figure 16 Cladisticanalyses including Caerorhachis :(a)Coates 1996; (b± d) Paton, Smithson & Clack1999. Abbreviationsas follows: AMN. amniotes;ANT. anthracosaurs;BAP. baphetids;COL. colosteids; DIA. diadectomorphs;LIS. lissamphibians;ˆ MIC. ˆ microsaurs;NEC. ˆnectrideans;SEY. ˆseymouria- morphs;ˆ TEM. temnospondyls.ˆ ˆ ˆ ˆ ˆ CAERORHACHIS BAIRDI 255 identi®cation of isolatedor disruptedtetrapod jaw fragments. ‘lepospondyls’,whereas diadectomorphs are the only Palaeo- Therelationships of various groups di Verfrom those of pre- zoicclade placed in thestem-group amniotes. Laurin’s matrix viousstudies (e.g. temnospondyls and ‘anthracosaurs’appear encompasses155 characters (47 scorable for Caerorhachis ) tobe polyphyletic). Ahlberg & Clack’smatrix consists of codedfor 43 livingand extinct taxa. Inclusion of Caerorhachis 26taxa and 50 characters. We notethat their character 48 and Whatcheeria doesnot alter the tree con® guration. We (presenceor absence of a postsplenialpit line) was coded as foundtwo trees (Fig. 17B; strict consensus), in which Caeror- 1(absence)in four taxa that lack a postsplenialaltogether hachis isnested between a temnospondylclade and a mono- (theircharacter 33), namely the nectrideans Sauropleura and phyletic ‘anthracosaur’clade. The position of Caerorhachis is Diploceraspis ,andthe amniotes Eocaptorhinus and Ophia- supportedby the following scorable characters: 3 (dermal codon.Thesetaxa should be coded as ?, inagreement with sculptureof shallow and widely spaced pits surrounded by Ahlberg& Clack’suse of questionmarks to indicateinapplic- smoothbony surface), 6 (jawarticulation near to occiput ablecharacters. We describethe results deriving from two out level),113 (capitulum of mid-presacral vertebrae articulates ofseveralcombinations of weighting and ordering regimes. tointercentrum and pleurocentrum of same segment; see also 5.5.1.All charactersunweighted and unordered. PAUP*yields commentsbelow), 119 (absence of uncinate processes), 143 968trees at 116 steps (C.I. 0:4779; R.l. 0:7122; R.C. (iliacblade with discrete dorsal and posterior ¯ anges),151 0:35).The proximal part of the ˆ strict consensus ˆ fails to provide ˆ (threephalanges in second pedal digit), 153 (® ve phalanges in resolutionfor all Devonian taxa, in contrast with the results fourthpedal digit). The position of Caerorhachis receives a obtainedby Ahlberg & Clack,who retrieved partial structure 22%bootstrap support. However, the 50% majority-rule con- inthe bottom half of theirconsensus tree. Whatcheeria is also sensustree shows Caerorhachis nestedbetween ( Proterogyrinus includedin such basal polytomy. More distally, Greererpeton , Archeria)andGephyrostegidae in a paraphyletic ‘anthra- Crassigyrinus , Megalocephalus and Caerorhachis are succes- cosaurs’.‡ These are placed between loxommatids and temno- sivelymore closely related to a poorlyresolved clade compris- spondyls. ing Cochleosaurus , Phonerpeton , Eoherpeton; Gephrostegus ; Ofall the characters (scorable and unscorable) optimised Microbrachis ; Discosauriscus ; (Pholiderpeton Proterogyrinus ); attheinternode leading to Caerorhachis andmore crownward (Balanerpeton Platyrhinops );andan unresolved ‡ amniote± tetrapods,those with the highest consistency indexes, 98 and nectrideanclade. ‡ The position of Caerorhachis receivesa 29% 113,are discussed here. Optimisation of character 98 assigns bootstrapsupport. We selectedarbitrarily one tree to examine state3 tothemorphological condition of centrain Caerorhachis optimisationof character changes supporting the position of (circularintercentra and pleurocentra). We decidedto code Caerorhachis .Thearrangement of taxa distal to the node Caerorhachis asunknown,because its vertebrae, although simi- subtending Caerorhachis features ‘anthracosaurs’as a clade larto those of ‘anthracosaurs’in the possession of enlarged consistingof ( Pholiderpeton Proterogyrinus )pairedwith pleurocentra,di Verfrom those of seymouriamorphs, West- (Eoherpeton Gephyrostegus ).‡Phonerpeton and Cochleosaurus lothiana,andamniotes s.s.Laurin’scoding for character 98 aresuccessively ‡ more closely related to sucha group.All these encompassesa largebut not exhaustive set of vertebral taxa join: (Balanerpeton Platyrhinops ); (Microbrachis Dis- morphologies;however, we didnot include an alternative cosauriscus ); (Sauropleura‡ (Ophiacodon (Eocaptorhinus‡ state for Caerorhachis inorder to retain as muchof the original Diploceraspis )).Theposition ‡ of Caerorhachis‡ issupported by ‡ matrixstructure as possible. Uncertainty characterises the thefollowing scorable characters: 7 (presenceof posterodorsal codingfor character 113, describing the modality of articula- processof posterior coronoid), 44 (presence of surangular tionof the mid presacral ribs to the vertebrae. We originally crest),45 (absenceof mandibularsensory canal). coded Caerorhachis as1,basedon comparisonswith Archeria, 5.5.2.Characters 36 and 46 ordered; all characters reweighted. Proterogyrinus ,certainseymouriamorphs and Westlothiana , in Withcharacters 36 and 46 ordered (referring, respectively, to whichthe capitulum of themid-presacral vertebrae articulates theposition of the centre of radiation on the prearticular, tothe intercentrum and pleurocentrum of the same vertebral andto the degree of enclosure of the mandibular sensory segment.However, recoding Caerorhachis asunknown for canal),a reweightedrun gives 90 trees (Fig. 17A; strict consen- character113 does not a Vectthe analysis. sus).The number, statistics and state changes of the characters supportingthe position of Caerorhachis inone of the minimal treesare almost identical to those of the unweighted and 6.Concludingremarks unorderedanalysis. The position of Caerorhachis receives a 35%bootstrapsupport. Work in progress on an expandedver- Variousfeatures of the lower jaw, palate, axial and appendi- sionof Ahlberg & Clack’smatrix (including such additional cularskeleton are consistent with the revised interpretation of charactersas proportions of infradentaries, extension of their Caerorhachis as a basal ‘anthracosaur’.Some of the characters mesiallaminae, size and position of Meckelianforamina, etc.) listedin the ‘Diagnosis’under various headings are also retrievestraditional taxonomic groups, including ‘anthra- observed(although not always in conjunction) in ‘anthra- cosaurs’and temnospondyls. These results, although prelimin- cosaurs’,seymouriamorphs, diadectomorphs and several pri- ary,suggest that the phylogenetic signal present in the original mitivecrown-group amniotes. Therefore, it is justi® able to matrixis in fact sensitive to character sampling. For example, regardthem as possible generalised amniote apomorphies at Caerorhachis canbe linked to ‘anthracosaurs’,and the diVerentlevels of inclusiveness(see discussion below). Analysis nectridean±amniote polytomy can be resolved in favour of ofcharacter distribution supporting the placement of Caeror- monophyleticnectrideans paired with monophyletic amniotes. hachis addsto the long-standing view that ‘anthracosaurs’ mayrepresent an early reptiliomorph radiation, despite their 5.6Laurin’s 1998a matrix veryprimitive nature (e.g. Panchen & Smithson1988; Smithson Laurin’s(1998a) database evolved from Carroll’s (1995), with et al.1994;Clack 1998b, 2001; Paton et al.1999).Furthermore, additionalcharacters. It covered all major Palaeozoic groups, detailedrecent investigations into the postcranium, braincase andwas intended to solvethe problem of lissamphibianorigins. andocciput morphology of seymouriamorphs and diadecto- Unlikeprevious analyses, it included representatives of Recent morphslend support to the amniote-like a nities of these lissamphibians,as well asprimitive crown-group amniotes. In clades(e.g. Sumida et al.1992;Sumida 1997; Berman 2000; Laurin’sphylogeny, lissamphibians are linked to a paraphyletic Berman et al.2000).However, a moreprecise assessment of 256 MARCELLORUTA ET AL.

Figure 17 Cladisticanalyses including Caerorhachis :(a)Ahlberg & Clack1998; (b) Laurin 1998a; abbreviations asinFigure 16.

theorder of character acquisition in the amniote stem-group Holmes1988) may represent an additionalamniote character, mustawait a large-scaleanalysis of Palaeozoictetrapod inter- giventheir presence in ‘anthracosaurs’(e.g. Carroll 1970; relationships(Clack 1994; Smithson 1994; Lombard & Bolt Panchen1985; Smithson 1985b; Holmes 1984, 1989; Clack 1995). 1987b,1998a), seymouriamorphs (e.g. White 1939; Klembara Amongthe least ambiguous features linking Caerorhachis to 1997),diadectomorphs (Berman et al.1998),and various basalreptiliomorphs are: subcircular depression on basalplate primitivecrown-group amniotes (e.g. Reisz 1981). However, ofparasphenoidwith thick peripheral rim; enlarged U-shaped theirpresence in Caerorhachis isdubious. The gastrocentrous pleurocentra;reduced dorsal iliac blade; L-shaped tarsal inter- (enlargedpleurocentrum) condition of the vertebral centra is medium.The depression on thebasal plate of the parasphenoid notunique to reptiliomorphs, because some microsaurs and appearsoccasionally in reptiliomorphs. Importantly, the devel- dissorophoidtemnospondyls reveal similar patterns. Interest- opmentof pronounced posterolateral ridges on the para- ingly,in the case of Paton et al.’s(1999) analysis microsaurs sphenoid(often ending in stout basal tubera; see Clack & appearas stem-groupamniotes; their gastrocentrous condition CAERORHACHIS BAIRDI 257 isthus congruent with that of ‘anthracosaurs’and crown-group saurus,seeRomer 1956), a shortridge merges with the dorsal amniotes.U-shaped pleurocentra occur in certain ‘anthraco- surfaceof the supra-acetabular buttress. Vaughn (1955) saurs’(e.g. Carroll 1970; Smithson 1985b), but no clear polarity drawsa faintline, corresponding in position to the ‘anthraco- isevident in the transformation of pleurocentra when other saur’ridge, on theilium of theearly amniote Araeoscelis . ‘anthracosaurs’(e.g. Holmes 1989), microsaurs and crown- Othercharacters of Caerorhachis areof dubiouspolarity or groupamniotes are considered (see Smithson et al. 1994). toopoorly known to be usedin a characteranalysis. Thus, the Reptiliomorphsshow di Verentdegrees of reduction of the size diVerencebetween vomerine and palatine/ ectopterygoid dorsaliliac blade, the presence of which is a primitivetetrapod fangsis also observed in other tetrapods (e.g. Edops, Eryops, feature(see Coates 1996). The ilium of Caerorhachis is surpris- Phonerpeton , Dendrerpeton ,etc.;Romer 1947; Dilkes 1990), inglyreminiscent of thatof variousbasal amniotes (e.g. Carroll andmay be growth-related.With regard to the shape of the pre- &Baird1972) in possessing a reduceddorsal blade. It isunlike frontal,the validity of this character rests on thedi cult, and theilia of e.g. Archeria and Proterogyrinus ,inwhich the dorsal admittedlyvery tentative interpretation of a highlyweathered bladeis pronounced. The morphology of the tarsal inter- skullroof element. If correctly interpreted, however, it suggests mediumwas discussed in detail by Lebedev & Coates(1995) anadditional baphetid-like feature in Caerorhachis , together andCoates (1996), who used it to link reptiliomorph ancestry withshape and position of the choanae, and morphology of tosuch early taxa as Tulerpeton,implyinga lateDevonian theanterior palatal region (see ‘Description’above). timeof divergence for the evolutionary split between lissam- Concerningthe palatal vacuities, these are widely distributed phibiansand amniotes (but see Clack 1998b, and Paton et al. amongPalaeozoic tetrapods (e.g. temnospondyls, colosteids, 1999).It is also observed in various ‘anthracosaurs’(e.g. somemicrosaurs and nectrideans, and various ‘anthra- Holmes1984, 1989), diadectomorphs and certain captorhinid cosaurs’).The de® nition of vacuities adopted here follows andprotorothyridid captorhinomorphs (e.g. Sumida 1997), as Milner& Sequeira(1994). It is based on their degree of curva- well asin some basal synapsids and diapsids (e.g. Reisz tureand relative width, but takes into account also the tri- 1981),but not in seymouriamorphs (e.g. Berman et al. 2000). radiateshape of the pterygoids and the elongate, parallel- Possibleadditional amniote characters of Caerorhachis , sidedcultriform process of the parasphenoid. None of these althoughnot uniquely derived, include: curved trunk ribs; conditionsis seen in Caerorhachis (althoughthe cultriform cervicalribs straight, morphologically distinct from trunk ribs processis poorly preserved). However, one or more of them andwith ¯ atteneddistal ends; rearmost part of mesial lamina characterisemost temnospondyls, some nectrideans, and at ofsplenialcloser to anteriormargin of adductor fossa than to leastone microsaur (Carroll & Gaskill1978; Milner 1993b; lowerjaw symphysis; transverse ridge sweeping from lateral Bossy& Milner1998). Thus, the simple presence of vacuities surfaceto ventral edge of posterior iliac process. Dorso- maybe of little signi® cance with regard to the supposed ventrallycurved trunk ribs are certainly primitive for amniotes ternnospondyla nities of Caerorhachis (contra Holmes2000). (e.g.Milner 1988, 1990; Coates 1996). However, they are Finally,it is hoped that our reassessment of Caerorhachis knownto occur in other groups (e.g. some microsaurs and makesthis taxon a suitablecandidate for inclusion in future, nectrideans,and large temnospondyls). Marked morphological comprehensivecladistic analyses of early tetrapod inter- diVerencebetween cervical/ pectoralribs and trunk ribs is also relationships. observedin some microsaurs (e.g. the genera Micraroter, Pantylus, Pelodosotis and Saxonerpeton ;Carroll& Gaskill 1978),and nectrideans (e.g. Diplocaulus ;Bossy& Milner 1998).The rearward extension of the mesial lamina of the 7.Acknowledgements splenial,closer to the anterior margin of the adductor fossa thanto the anterior end of the mandible, is found in DrsAnne Warren (La Trobe University, Melbourne) and baphetids, ‘anthracosaurs’,seymouriamorphs, diadectomorphs AdamYates (University of Bristol) provided stimulating and andamniotes s.s.(White1939; Bystrow 1944; Romer 1956; constructivecriticism. Dr AndrewSmith (The Natural History Beaumont1977; Holmes 1984, 1989; Smithson 1986; Clack Museum,London) o Veredhelpful editorial comments. Pro- 1987b;Klembara 1997; Ahlberg & Clack1998). It is also fessorRobert Carroll (McGill University, Montreal) and presentin some microsaurs (e.g. Euryodus, Microbrachis , DrAngela Milner (The Natural History Museum, London) Pantylus;Carroll& Gaskill1978), and in all nectrideans in lentus casts of Caerorhachis .DrsJenny Clack (University of whichthe mesial aspect of thelower jaw is known in su cient Cambridge)and Tim Smithson (Aylesbury College, Bucking- detail(e.g. Sauropleura,Batrachiderpeton, Diploceraspis ; Bossy hamshire)exchanged much useful information. Dr Clack &Milner1998). Importantly, this character is not linked to the kindlysupplied a copyof the Casineria matrixand a pre- occurrenceof a postsplenial,or to the relative length of the publicationdraft of her Eucritta redescription.This work is adductorfossa. Finally, a transverseridge sweeping from fundedby BBSRC AdvancedResearch Fellowship no. 31/ thelateral surface to the ventral margin of the posterior iliac AF/13042awarded to MichaelI. Coates. processoccurs in ‘anthracosaurs’(e.g. Carroll 1970; Holmes 1984,1989; Smithson 1985b), seymouriamorphs (e.g. White 1939;Klembara & Bartik2000), diadectomorphs (e.g. Romer 1956;Berman & Sumida1990; Berman et al.1998),and (spor- 8.Abbreviationsused in ® gures adically)basal amniotes (e.g. Romer 1956; Vaughn 1955). The temnospondyls Eryops and Dendrerpeton possessfaint iliac a.ac.nÐanterior acetabular notch ridges(Romer 1922; Holmes et al.1998),although these are acÐacetabulum almostcertainly examples of convergence with ‘anthracosaurs’. add.crÐadductor crest In Dendrerpeton ,thecourse and orientation of the iliac ridge ad.foÐadductor fossa resembleclosely those of Caerorhachis ,althoughthe ridge of ANGÐangular Dendrerpeton isnot accompanied by muscle scars. The iliac ant.CORÐanterior coronoid ridgeis modi® ed in seymouriamorphs and diadectomorphs, ant.M.fÐanterior Meckelian foramen whereit delimits a lateraliliac shelf (e.g. Romer 1956; ARTÐarticular

Klembara& Bartik2000). In some synapsids (e.g. Edapho- c1Ðcentrale1 258 MARCELLORUTA ET AL. c2Ðcentrale2 9.Appendix.Coding for Caerorhachis in diVerent c3Ðcentrale3 data sets chÐchoana DEÐdentary Carroll1995 dibÐdorsal iliac blade 1??????1?? ??01? ???11 00000 00112 1??00 01020 11?10 000?1 ECPTÐectopterygoid ?21?????2? ?1001 300?0 011?? ????? ??121 ???1? ????3 ?201? FBLЮbulare 210101(12)012 1?1(12)? ????? ????? ??(12)11 (012)???? ????? FBL.faЮbulare facet ?????(012)??2? ????1 ???11 11?11 1111(12) 11112 (12)411? 1111 ®.faЮbular facet FRÐfrontal Coates 1996 fun.excÐfunnel-shaped excavation (posterolateral wall of ?2???001?1 (12)(12)12? ????1 0??11 11?(12)1 0(12)02? ??1?0 1?1?? choana) ?11?11???? 10?1? 1?1?1 ??10? 1???0 hsÐhaemal spine icÐintercentrum Clack 1998a ico.grÐintercondylar groove ??200????? ??0?? ????? ?00?0 0???? ??2(01)0 00??1 10??? 01211 ILÐilium 1???011100 1110? 01111 ?01?? ????? ?111? 1?1 INTÐintermedium INT.fa.fÐfacet for intermedium on ®bula Clack 1998b INT.fa.tÐfacet for intermedium on tibia ?1??1??00? ????? ????? ????0 0(01)0?? ????2 (01)?000 ???11 0??11 in.trÐinternal trochanter 2110110111 00111 010?? 01111 ?201? ????? ?011? ????? ?0??? ISCHÐischium ??11000?11 1 ITÐintertemporal MAXÐmaxilla Ahlberg& Clack1998 mes.lam.ANGÐmesial lamina of angular 0110101131 11111 ???11 01101 00001 01010 ?0010 00111 ???01 mes.lam.POSPLÐmesial lamina of postsplenial mes.lam.SPLÐ mesial lamina of splenial Laurin1998a mid.CORÐmiddle coronoid (02)?20?110?? (01)0000 000?0 00000 ?00?? 20000 02020 10001 mid.M.fÐmiddle Meckelian foramen 0010000100 01000 ???0(23) 1100? 1??0? ???20 00000 00100 msmÐmesial shelf of maxilla 0100001??? ?001? ????? 1?101 1110(01) 00??? ????0 ????? ??01? mttÐmetatarsal (23)?1(12)?(12)0002 10221 NAÐnasal n.arÐneural arch Paton,Smithson & Clack1999 PAÐparietal 1?????00?? ????? ?1121 10000 ?0??? (01)???2 (01)??00 ??0?? ?1011 p.ac.nÐposterior acetabular notch 0011101??0 01111 ?201? ????? ?0?01 ????? 0???? ?0??? 001?? PALÐpalatine 11(01)(01)(456)00??1 1 PASÐparasphenoid pcÐpleurocentrum pdipÐposterodorsal iliac process 10.References phÐphalanx POÐpostorbital Ahlberg,P .E. 1998.Postcranial stem tetrapod remains from the POFRÐpostfrontal Devonianof Scat Craig, Morayshire, Scotland. ZoologicalJournal ofthe Linnean Society 122, 99±141. POSPLÐpostsplenial Ahlberg.P. E.&Clack,J. A.1998.Lower jaws, lower tetrapodsÐ post.CORÐposterior coronoid areviewbased on theDevonian genus Acanthostega.Transactions post.M.fÐposterior Meckelian foramen ofthe Royal Society of Edinburgh:Earth Sciences 89, 11±46. PPÐpostparietal Ahlberg,P. E.&Milner, A.R.1994.The origin and early diversi® ca- PREARTÐprearticular tionof tetrapods. Nature 368, 507±14. Andrews,S. M.&Brand,P. J.1991.Postscript: confoundedironstones PSYMÐparasymphysial plate inthe Limestone Coal Group of Loanhead, Midlothian. Transac- PTÐpterygoid tionsof theRoyal Society of Edinburgh:Earth Sciences 82, 273±5. PUÐpubis Beaumont,E. H.1977.Cranial morphology of the Loxommatidae QJÐquadratojugal (Amphibia:Labyrinthodontia). PhilosophicalTransactions of the RoyalSociety of London B280, 29±101. rbÐround bosses Beaumont,E. H.&Smithson,T. R.1998.The cranial morphology and r.rÐrugose ridge for ¯ exormuscle attachment on tibia relationshipsof the aberrant Carboniferous amphibian Spathi- sabÐsupra-acetabular buttress cephalusmirus Watson. ZoologicalJournal of theLinnean Society SPLÐsplenial 122, 187±209. SQÐsquamosal Berman,D. S.2000.Origin and early evolution of the amniote occiput. Journalof Paleontology 74, 938±56. STÐsupratemporal Berman,D. S.,Reisz, R. R.&Eberth,D. A.1985. Ecolsoniacutlerensis , SURANGÐsurangular anearly Permian dissorophid amphibian from the Cutler Forma- sur.creÐsurangular crest tionof north-central New Mexico. Circularof the New Mexico symÐsymphysis Bureauof Minesand Mineral Resources 191, 1±31. Berman,D. S.,Sumida,S. S.&Martens,T. 1998. Diadectes (Diadecto- TAÐtabular morpha:Diadectidae) from the early Permian of centralGermany, TBL+c4Ðtibiale+centrale 4 with descriptionof anewspecies. Annalsof CarnegieMuseum 67, TBL.faÐtibiale facet 53±93. ti.faÐtibial facet Berman,D. S.,Henrici, A. C., Sumida,S. S.&Martens,T. 2000. tr.pel.rÐtransverse pelvic ridge Redescriptionof Seymouriasanjuanensis (Seymouriamorpha) fromthe Lower Permianof Germany based on complete, mature tslÐtarsal specimenswith adiscussionof paleoecology of the Bromacker tu.istrÐtuberosity for insertion of ischiotrochantericus muscle localityassemblage. Journalof Vertebrate Paleontology 20, 253± VOMÐvomer 68. CAERORHACHIS BAIRDI 259

Berman,D. S.&Sumida,S. S.1990.A newspecies of Limnoscelis Dilkes,D. W.1990.A newtrematopsid amphibian (Temnospondyli: (Amphibia,Diadectomorpha) from thelate Pennsylvanian Dissorophoidea)from the Lower Permianof Texas. Journal of Sangrede Cristo Formationof central Colorado. Annals of VertebratePaleontology 10, 222±43. CarnegieMuseum 59, 303±41. Fox,R. C., Campbell,K. S.W., Barwick,R. E.&Long,J. A.1995.A Bolt, J.R.1969.Lissamphibian origins: possible protolissamphibian newosteolepiform ® sh fromthe Lower CarboniferousRaymond fromthe Lower Permianof Oklahoma. Science 166, 888±91. Formation,Drummond Basin, Queensland. Memoirsof the Bolt, J.R.1979. Amphibamusgrandiceps asa juveniledissorophid: QueenslandMuseum 38, 97±221. evidenceand implications. In Nitecki,M. H.(ed.). Mazon Creek Gardiner,B. G.1984.The relationships of the palaeoniscid ® shes, a Fossils,529±63. New York: Academic Press. reviewbased on new specimens of Mimia and Moythomasia Bolt, J.R.1991.Lissamphibian origins. In Schultze,H.-P. &Trueb,L. fromthe Upper Devonian of Western Australia. Bulletin of (eds) Originsof thehigher groups of tetrapods: controversy and con- theBritish Museum(Natural History): Geology Series 37, 173± sensus,194±222. Ithaca: Cornell University Press. 428. Bossy, A.K.&Milner, A.C.1998.Order Nectridea Miall, 1875. In Gauthier,J. A.,Kluge,A. G.&Rowe,T. 1988.The early evolution of Carroll, R.L., Bossy,K. A.,Milner, A.C., Andrews,S. M.& theAmniota. In Benton,M. J.(ed.) Thephylogeny and classi® - Wellstead,C. F.(eds) Teil1, Lepospondyli. Handbuch der Pala Èo- cationof the tetrapods , 1:Amphibians, Reptiles, Birds , 103±55. herpetologie ,73±212. Munich: Pfeil. Oxford:Clarendon Press. Busbey,A. B.1995.The structural consequences of skull ¯attening Godfrey,S. J.1989.The postcranial skeletal anatomy of theCarboni- incrocodilians. In Thomason,J. J.(ed.) Functionalmorphology feroustetrapod Greererpetonburkemorani. Philosophical Transac- invertebrate paleontology ,173±92. Cambridge: Cambridge tionsof theRoyal Society of London B323, 75±133. UniversityPress. Godfrey,S. J.,Fiorillo, A.R.&Carroll, R.L.1987.A newlydiscovered Bystrow, A.P.1944.On Kotlassiaprima (Amalitsky). Bulletinof the skull ofthe temnospondyl amphibian Dendrerpetonacadianum GeologicalSociety of America 55, 379±416. Owen. CanadianJournal of EarthSciences 24, 796±805. Carroll. R.L.1970.The ancestry of reptiles. PhilosophicalTransactions Goodrich,E. S.1930. Studieson the structure and development of ofthe Royal Society of London B257, 267±308. vertebrates .London:Macmillan & Co. Carroll, R.L.1988. Vertebratepaleontology and evolution . New York: Harland,W. B., Armstrong,R. L., Cox,A. V., Craig,L. E., Smith, Freeman. A.G.&Smith,D. G.1989. Ageologictime scale 1989 .Cambridge: Carroll, R.L.1991. Batropetes fromthe Lower Permian of EuropeÐ CambridgeUniversity Press. amicrosaur,not a reptile. Journalof VertebratePaleontology 11, Heaton,M. J.1979.Cranial anatomy of primitivecaptorhinid reptiles 229±42. fromthe Late Pennsylvanian and Early Permian, Oklahoma and Carroll, R.L.1995.Problems of the phylogenetic analysis of Paleozoic Texas. Bulletinof theOklahoma Geogical Suevey 127, 1±80. choanates.In Arsenault,M. LelieÁvre,H. &Janvier,P. (eds) Studies Heaton,M. J.1980.The Cotylosauria: a reconsiderationof a groupof onearly vertebrates: (VIIth International Symposium, Parc de archaictetrapods. In Panchen,A. L.(ed.) Theterrestrial environ- Miguasha,Quebec). Bulletin du Museum National d’Histoire mentand the origin of landvertebrates ,497±55l. London: Academic Naturelle,Paris (Ser.4) 17 (C), 38±445. Press. Carroll. R.L.1996.Elongate early tetrapods: why ‘Lizzie’had little Holmes,R. B.1984.The Carboniferous amphibian Proterogyrinus limbs. In Milner, A.R.(ed.) Studieson Carboniferous and Permian scheelei Romer,and the early evolution of tetrapods. Philosophical vertebrates.Special Papers in Palaeontology 52,139±48. Transactionsof theRoyal Society of London B306, 431±527. Carroll, R.L., Bybee.P. &Tidwell, W. D.1991.The oldest microsaur Holmes,R. B.1989.The skull andaxial skeleton of the Lower Permian (Amphibia). Journalof Paleontology 65, 314±22. anthracosauroidamphibian Archeriacrassidisca Cope. Palaeonto- Carroll, R.L., Bossy, K.A.,Milner, A.C., Andrews,S. M. &Well- graphica 207, 161±206. stead,C. F.1998. Teil1, Lepospondyli. Handbuch der Palao- Holmes,R. B.2000.Palaeozoic temnospondyls. In Heatwole,H. E.& herpetologie .Munich:Pfeil. Carroll, R.L.(eds) AmphibianBiology 4: Palaeontology , 1081±120. Carroll, R.L.&Baird,D. 1972.Carboniferous stem-reptiles of ChippingNorton: Surrey Beatty & Sons. theFamily Romeriidae. Bulletinof the Museum of Comparative Holmes,R. B., Carroll, R.L.&Reisz,R. R.1998.The ® rst articulated Zoology 143, 321±64. skeleton of Dendrerpetonacadianum (Temnospondyli,Dendrer- Carroll, R.L.&Gaskill.P. 1978.The Order Microsauria. Memoirs of petontidae)from the lower Pennsylvanianlocality of Joggins, theAmerican Philosophical Society 126, 1±211. NovaScotia, and a reviewof its relationships. Journal of Clack,J. A.1987a.Two newspecimens of Anthracosaurus (Amphibia: VertebratePaleontology 18, 64±79. Anthracosauria)from the Northumberland Coal Measures. Holmes,R. B.&Carroll, R.L.1977.A temnospondylamphibian from Palaeontology 30, 15±26. theMississippian ofScotland. Bulletinof theMuseum of Compara- Clack.J. A.1987b. Pholiderpetonscutigerum Huxley,an amphibian tiveZoology 147, 489±511. fromthe Yorkshire Coal Measures. PhilosophicalTransactions of Hook,R. W.1983. Colosteusscutellatus (Newberry),a primitivetem- theRoyal Society of London B318, 1±107. nospondylamphibian from the Middle Pennsylvanian of Linton, Clack,J. A.1994. Silvanerpetonmiripedes ,anewanthracosauroid from Ohio. AmericanMuseum Novitates 2770, 1±41. theVise ÂanofEast Kirkton, West Lothian,Scotland. Transactions Jarvik,E. 1980. Basicstructure and evolution of vertebrates 1, 1±591. ofthe Royal Society of Edinburgh: Earth Sciences 84, 369±76. Londonand New York: Academic Press. Clack,J. A.1998a.The Scottish Carboniferous tetrapod Crassigyrinus Jarvik,E. 1996.The Devonian tetrapod Ichthyostega.Fossils &Strata scotcus (Lydekker)Ðcranial anatomy and relationships. Trans- 40, 1±206. actionsof theRoyal Society of Edinburgh: Earth Sciences 88 (for Klembara,J. 1997.The cranial anatomy of Discosauriscus Kuhn, a 1997),127± 42. seymouriamorphtetrapod from the Lower Permianof the Bosko- Clack,J. A.1998b.A newEarly Carboniferous tetrapod with a meÂlange viceFurrow (CzechRepublic ). PhilosophicalTransactions of the ofcrown-group characters. Nature 394, 66±9. RoyalSociety of London B352, 257±302. Clack,J. A.2001. Eucrittamelanolimnetes fromthe Early Carbonifer- Klembara,J. &Bartik,I. 2000.The postcranial skeleton of Discosaur- ousof Scotland, a stemtetrapod showing a mosaicof characteris- iscus Kuhn,a seymouriamorphtetrapod from the Lower Permian tics. Transactionsof the Royal Society of Edinburgh: Earth Sciences ofthe Boskovice Furrow (CzechRepublic ). Transactionsof the 92, 75±95. RoyalSociety of Edinburgh:Earth Sciences 90 (for 1999),287± 316. Clack,J. A.&Holmes,R. 1988.The braincase of the anthracosaur Laurin,M. 1998a.The importance of global parsimony and historical Archeriacrassidisca with commentson the interrelationships of biasin understanding tetrapod evolution. Part I. Systematics, primitivetetrapods. Palaeontology 31, 85±107. middleear evolution, and jaw suspension. Annalesdes Sciences Coates,M. I.1996.The Devonian tetrapod Acanthostegagunnari Naturelles,Zoologie 1, 1±42. Jarvik:postcranial anatomy, basal tetrapod interrelationships Laurin,M. 1998b.The importance of global parsimony and historical andpatterns of skeletal evolution. Transactionsof the Royal biasin understanding tetrapod evolution. Part II. Vertebral Societyof Edinburgh: Earth Sciences 87, 363±421. centrum,costal ventilation, and paedomorphosis. Annales des Coates,M. I., Milner, A.R.&Ruta,M. 2000.Early tetrapod evolution. SciencesNaturelles, Zoologie 19, 99±114. Trendsin Ecology and Evolution 15, 327±8. Laurin,M. 1998c.A reevaluationof the origin of pentadactyly. Daeschler,E. B.2000.Early tetrapod jaws fromthe late Devonian of Evolution 52, 1476±82. Pennsylvania,USA. Journalof Paleontology 74, 301±8. Laurin,M., Girondot,M. &deRicqle Ás, A.2000a.Early tetrapod Daeschler,E. B., Shubin,N., Thomson,K. S.&Amaral,W. W.1994. evolution. Trendsin Ecology and Evolution 15, 118±23. ADevoniantetrapod from North America. Science 265, 639± Laurin,M., Girondot,M. &deRicqles Á ,A.2000b.Reply. Trends in 42. Ecologyand Evolution 15, 328. 260 MARCELLORUTA ET AL.

Laurin,M. &Reisz,R. R.1997.A newperspective on tetrapod phylo- Paton,R. L., Smithson,T. R.&Clack,J. A.1999.An amniote-like geny. In Sumida,S. S.&Martin, K.L.M. (eds) Amnioteorigins: skeletonfrom the Early Carboniferous of Scotland. Nature 398, completingthe transition to land ,9±59. London: Academic Press. 508±13. Laurin,M. &Reisz,R. R.1999.A newstudy of Solenodonsaurus Pough,F. H., Andrews,R. M., Cadle,J. E., Crump, M.L., Savitzky, janenschi,anda reconsiderationof amniote origins and stego- A.H.&Wells, K.D2000. Herpetology ,2ndedn. Upper Saddle cephalianevolution. CanadianJournal of Earth Sciences 36, River:Prentice Hall. 1239±55. Reisz,R. R.1981.A diapsidreptile from the Permian of Kansas. Lebedev,O. A.&Clack,J. A.1993.Upper Devonian tetrapods from SpecialPublications of theMuseum of Natural History 7, 1±74. Andreyevka,Tula region, Russia. Palaeontology 36, 721±34. Romer,A. S.1922.The locomotor apparatus of certainprimitive and Lebedev,O. A.&Coates,M. I.1995.The postcranial skeleton of the mammal-likereptiles. Bulletinof theAmerican Museum of Natural Devoniantetrapod Tulerpetoncurtum Lebedev. ZoologicalJournal History 46, 517±606. ofthe Linnean Society 114, 307±48. Romer,A. S.1947.Review of the Labyrinthodontia. Bulletinof the Lombard,R. E.&Bolt, J.R.1995.A newprimitive tetrapod Whatch- Museumof ComparativeZoology 99, 1±368. eeriadeltae fromthe Lower Carboniferousof Iowa. Palaeontology Romer,A. S.1956. Osteologyof the reptiles .Chicago:University of 38, 471±94. ChicagoPress. Long,J. A.,Barwick,R. E.&Campbell,K. S.W.1997.Osteology and Romer,A. S.1957.The appendicular skeleton of the Permian embol- functionalmorphology of the osteolepiform ® sh Gogonasus omerousamphibian Archeria.Contributions of the Museum of andrewsae Long,1985, from the Upper Devonian Gogo Forma- Geologyof the University of Michigan 13, 103±59. tion,Western Australia. Recordsof the Western Australian Romer,A. S.1966. Vertebratepaleontology .Chicago:University of Museum,Supplement 53, 1±89. ChicagoPress. Maddison,W. P.&Maddison,D. R.1992. MacClade:Analysis of Romer,A. S.&Parsons, T.S.1977. Thevertebrate body , 5th edn. phylogenyand character evolution .Version 3.0.1.Sunderland, Philadelphia:W. B.Saunders. Massachusetts:Sinauer Associates. Milner, A.C.1980.A review Romer,A. S.&Witter, R.V.1942. Edops,aprimitiverhachitomous ofthe Nectridea (Amphibia). In Panchen,A. L.(ed.) Theterrestrial amphibianfrom the Texas red beds. Journalof Geology 50, environmentand the origin of land vertebrates ,377±405. London: 925±60. AcademicPress. SaÈve-SoÈderbergh,G. 1934.Some points of viewconcerning the evolu- Milner, A.C.&Lindsay,W. 1998.Postcranial remains of Baphetes tionof the vertebrates and the classi® cation of this group. Arkiv andtheir bearing on the relationships of the Baphetidae fuÈr Zoologi 26, 1±20. ( Loxommatidae). ZoologicalJournal of the Linnean Society Schoch,R. R.&Milner, A.R.2000. Teil3B, Stereospondyli. Handbuch 122ˆ , 211±35. der PalaÈoherpetologie .Munich:Pfeil. Milner, A.R.1980.The temnospondyl amphibian Dendrerpeton Sequeira,S. E.K.1996.A cochleosauridamphibian from the Upper fromthe Upper Carboniferous of Ireland. Palaeontology 23, Carboniferousof Ireland. In Milner, A.R.(ed.) Studieson Carbon- 125±41. iferousand Permian vertebrates. Special Papers in Palaeontology Milner, A.R.1988.The relationships and origin of living amphibians. 52, 65±80. In Benton,M. J.(ed.) Thephylogeny and classi® cation of the tetra- Sequeira,S. E.K&Milner, A.R.1993.The temnospondyl amphibian pods,1: Amphibians, Reptiles, Birds ,59±102. Oxford: Clarendon Capetus fromthe Upper Carboniferous of Ny y ryÆ any, Czech Press. Republic. Palaeontology 36, 657±80. Milner, A.R.1990.The radiations of temnospondyl amphibians. Smithson,T. R.1980.An earlytetrapod fauna from the Namurian of In Taylor,P. D.&Larwood,G. P(eds) Majorevolutionary Scotland. In Panchen,A. L.(ed.) Theterrestrial environment radiations,321±49. Oxford: Clarendon Press. andthe origin of land vertebrates ,407±38. London: Academic Milner, A.R.1993a.Amphibian-grade Tetrapoda. In Benton,M. J. Press. (ed.) TheFossil Record2 ,665±79. London: Chapman & Hall. Smithson,T. R.1982.The cranial morphology of Greererpeton Milner, A.R.1993b.The Paleozoic relatives of lissamphibians. burkemorani Romer(Amphibia: Temnospondyli). Zoological HerpetologicalMonographs 7, 8±27. Journalof theLinnean Society 76, 29±90. Milner, A.R.,Smithson,T. R.,Milner, A.C., Coates,M. I. &Rolfe, Smithson,T. R.1985a.Scottish Carboniferous amphibian localities. W. D.I.1986.The search for earlytetrapods. ModernGeology ScottishJournal of Geology 21, 123±42. 10, 1±28. Smithson,T. R.1985b.The morphology and relationships of the Milner, A.R.&Sequeira,S. E.K.1994.The temnospondyl amphibians Carboniferousamphibian Eoherpetonwatsoni Panchen. Zoological fromthe Vise ÂanofEast Kirkton, West Lothian,Scotland. Trans- Journalof theLinnean Society 85, 317±410. actionsof theRoyal Society of Edinburgh: Earth Sciences 84 (for Smithson,T. R.1986.A newanthracosaur amphibian from the 1993),331± 61. Carboniferousof Scotland. Palaeontology 29, 603±28. Milner, A.R.&Sequeira,S. E.K.1998.A cochleosauridtemnospondyl Smithson,T. R.1994. Eldeceeonrolfei ,anewreptiliomorph from the amphibianfrom the 6 MiddlePennsylvanian of Linton, Ohio, ViseÂanof EastKirkton, West Lothian,Scotland. Transactionsof U.S.A. ZoologicalJournal of the Linnean Society 122, 261±90. theRoyal Society of Edinburgh: Earth Sciences 84 (for 1993), Moulton,J. M.1974.A descriptionof thevertebral column of Eryops 377±82. basedon the notes and drawings of A. S.Romer. Breviora 428, Smithson,T. R.,Carroll, R.L., Panchen,A. L.&Andrews,S. M.1994. 1±44. Westlothianalizziae fromthe Vise  anof East Kirkton, West Panchen,A. L.1973.On Crassigyrinusscoticus Watson,a primitive Lothian,Scotland, and the amniote stem. Transactionsof the amphibianfrom the Lower Carboniferousof Scotland. Palae- RoyalSociety of Edinburgh: Earth Sciences 84 (for 1993),383± ontology 16, 179±93. 412. Panchen,A. L.1977.On Anthracosaurusrusselli Huxley(Amphibia: Sullivan,C., Reisz,R. R.&May,W. J.2000.Large dissorophoid Labyrinthodontia)and the family Anthracosauridae. Philo- skeletalelements from the Lower PermianRichards Spur ® ssures, sophicalTransactions of the Royal Society of London B279, 447± Oklahoma,and their paleoecological implications. Journal of 512. VertebratePaleontology 20, 456±61. Panchen,A. L.1980.The origin and relationships of the anthracosaur Sumida,S. S.1997.Locomotor features of taxa spanning the origin of Amphibiafrom the late Palaeozoic. In Panchen,A. L.(ed.) The amniotes. In Sumida,S. S.&Martin, K.L.M.(eds) Amniote terrestrial environmentand the origin of land vertebrates , 319±50. origins:completing the transition to land ,353±98. London: London:Academic Press. AcademicPress. Panchen,A. L.1985.On the amphibian Crassigyrinusscoticus Watson Sumida,S. S.,Lombard, R. E.&Berman,D. S.1992.Morphology of fromthe Carboniferous of Scotland. PhilosophicalTransactions of theatlas-axis complex of the late Palaeozoic tetrapod suborders theRoyal Society of London B309, 505±68. Diadectomorphaand Seymouriamorpha. PhilosophicalTrans- Panchen,A. L.&Smithson,T. R.1987.Character diagnosis, fossils, actionsof theRoyal Society of London 336, 259±73. andthe origin of tetrapods. BiologicalReviews 62, 341±438. SwoVord,D. L.1993. PAUP:Phylogenetic analysis using parsimony . Panchen,A. L.&Smithson,T. R.1988.The relationships of early Version 3.1.1.Champaign, Illinois: Illinois NaturalHistory tetrapods. In Benton,M. J.(ed.) Thephylogeny and classi® cation Survey. ofthe tetrapods 1,1±32. Oxford: Clarendon Press. SwoVord,D. L.2000. PAUP*:Phylogenetic analysis using parsimony Panchen,A. L.&Smithson,T. R.1990.The pelvic girdle and hind limb (*andother methods) .Version 4.0b4a.Sunderland, Massachusetts: of Crassigyrinusscoticus (Lydekker)from the Scottish Carbon- SinauerAssociates. iferousand the origin of the tetrapod pelvic skeleton. Transactions Vaughn,P. P.1955.The Permian reptile Araeoscelis restudied. Bulletin ofthe Royal Society of Edinburgh: Earth Sciences 81, 31±44. ofthe Museum of Comparative Zoology 113, 305±467. CAERORHACHIS BAIRDI 261

Wellstead,C. F.1991.Taxonomic revision of the Lysorophia, Permo- vertebrates. Bulletinof the Geological Society of America 21, 249± Carboniferouslepospondyl amphibians. Bulletinof theAmerican 84. Museumof NaturalHistory 209, 1±90. Yates,A. M.&Warren,A. A.2000.The phylogeny of the ‘higher’ White,T. E.1939.Osteology of Seymouriabaylorensis (Broili). Bulletin temnospondyls(Vertebrata: Choanata) and its implicationsfor ofthe Museum of Comparative Zoology 85, 325±409. themonophyly and origins of the Stereospondyli. Zoological Williston, S.W.1910. Cacops,Desmospondylus :newgenera of Permian Journalof theLinnean Society 128, 77±121.

MARCELLO RUTA*and MICHAEL I.COATES*,Department of Biology,University College London,Gower Street, London WC1E 6HT,UK. *Presentaddress: Department of OrganismalBiology and Anatomy, University of Chicago, 1027East 57th Street, Chicago, IL 60637,U.S.A. e-mail:[email protected] u;[email protected]. edu ANDREWR. MILNER,Department of Biology,Birkbeck College, Malet Street, London WC1E 7HX,UK. e-mail:[email protected]

MSreceived15 December2000. Accepted for publication 29 June2001.