Middle-Upper Ordovician conodonts from the Ffairfach and Golden Grove groups in South Wales, United Kingdom

ABSTRACT The conodont fauna of the reference succession of the regional British Llandeilian Stage of the Llanvirn Series was first described in a classical study by Rhodes more than 65 years ago using single element (form) taxonomy. Although several subsequent authors have recorded a substantial number of conodont taxa from the Llandeilo area, the present study is the first to present a modern taxonomic review of these late Darriwilian-early Sandbian faunas that include approximately 20 multielement species. Most prominent are representatives of Amorphognathus, Baltoniodus, Eoplacognathus, and Plectodina. The study faunas have their own biogeographical character. The distinctive genera Complexodus, Protopanderodus, and Pygodus, which are common in coeval Baltoscandic faunas, are not present, but the occurrence of Amorphognathus, Baltoniodus, and Eoplacognathus provides a link to age equivalent Baltoscandic faunas. The presence of abundant specimens of Plectodina and less common representatives of Erismodus and Icriodella are reminiscent of North American Midcontinent faunas. This type of faunal assemblage is in some respects similar to those of the early Caradoc Series of the Welsh Borderland. Biostratigraphically diagnostic species indicate that the Llandeilo study succession ranges from the Eoplacognathus lindstroemi Subzone of the Pygodus serra Zone to the Baltoniodus variabilis Subzone of the Amorphognathus tvaerensis Zone.


Introduction
There is no doubt that the beginning of the post-pioneer period of research on Lower Palaeozoic conodonts in not only the United Kingdom but also in the entire Europe is marked by the publication of Rhodes' (1953) monograph on the Ordovician and Silurian conodonts from Wales and the Welsh Borderland. One important portion of this classical work is the description of a conodont fauna from the Llandeilo Limestone. This unit, which is well known in British Lower Palaeozoic geology, was named (Murchison 1839) for the small town of Llandeilo in South Wales (Figure 1), where there are many outcrops of the unit. The term Llandeilo became internationally known as a standard series unit in the British regional series classification of the Ordovician System (e.g., Williams et al. 1972). What Rhodes (1953) referred to as the Llandeilo Limestone is now known as the Llandeilo Group, which is subdivided into three formations, the Lower, Middle, and Upper Llandeilo Flags ( Figure 2).
As was the practice at that time, Rhodes (1953) used single element (form) taxonomy in his description of this conodont fauna. He recognised 16 species, among which five were new and one not identified at the species level. A restudy of the Rhodes collection along with additional samples led Bergström (1964) to identify 23 form species from the Llandeilo Group, three of which were not identified at the species level. Some further observations on the Llandeilo Group conodonts were published by Bergström (1971). Part of this fauna was reported in terms of multielement taxonomy by Bergström and Orchard (1985) but a more comprehensive account was published by Bergström et al. (1987). Their account was based on many additional samples not only from the Llandeilo area but also from coeval strata in the Narberth area some 60 km west-southwest of Llandeilo. The study of Bergström et al. (1987) was mainly of biostratigraphical nature and did not include any formal taxonomic descriptions. Whereas a considerable amount of taxonomic work has been carried out on slightly younger conodont faunas in Wales and the Welsh Borderland (e.g., Lindström 1959;Orchard 1980;Savage and Bassett 1985;Ferretti et al. 2014a;Bergström and Ferretti 2018), the conodont faunas of the Llandeilo Group and the slightly older Ffairfach Group have never been formally described in terms of modern conodont taxonomy. The purpose of the present study is to present an illustrated taxonomic description of these important conodont faunas and assess their significance in terms of also biostratigraphy, biogeography and evolution of some taxa.

Current stratigraphical classification of the study interval and conodont sample levels
During the more than half a century since the publication of the Rhodes (1953) classic paper, there have been very significant changes in the stratigraphical classification of his study unit (Figure 2), and a brief review of this reclassification is needed for placing these faunas in a modern stratigraphical framework. What was formerly known as the Llandeilo Limestone is now classified as three formations, the Lower, Middle, and Upper Llandeilo Flags, which constitute the Golden Grove Group (cf. Sutton et al. 1999;Fortey et al. 2000). This group rests on strata of the Ffairfach Group, which is subdivided into five formations (Figure 2). For a long time, the Llandeilo Series was taken to correspond to the Glyptograptus (now Hustedograptus) teretiusculus graptolite Zone (e.g., Fortey et al. 1995) but studies of conodonts (Bergström 1971;Bergström et al. 1987) and shelly fossils (Addison in Williams et al. 1972) showed that only part of the Lower Llandeilo Flags corresponds to that graptolite zone, the Middle and Upper Llandeilo Flags formations being coeval to the lower part of the Nemagraptus gracilis graptolite Zone, which by definition marks the lower part of the regional Caradoc Series. Furthermore, the regional British Llandeilo Series, which has its reference area at Llandeilo, has been reduced to the Llandeilian Stage of the Llanvirn Series (Fortey et al. 1995(Fortey et al. , 2000. The top of the revised Llanvirn Series is now defined as the base of the Nemagraptus gracilis graptolite Zone, and the base of the current Llandeilian Stage as the base of the Hustedograptus teretiusculus graptolite Zone. Unfortunately, because biostratigraphically diagnostic graptolites are absent in this stratigraphical interval in the Llandeilo region, neither the precise level of the top, nor that of the  (SN 592198). This is Locality 1 of Rhodes (1953). b, detail of Site 1 showing conodont sample levels; modified after Williams (1952) and Bergström et al. (1987). c, location of Sites 3-5 in the Dynevor Park. Site 3 is along dirt road 0.5 km west south-west of Dynevor Castle (SN 608223). This is Stop II-10 of Bassett et al. (1974). Site 4 is disused quarry in the Upper Llandeilo Flags 0.25 km north-east of Dynevor Castle (SN 616227). This is Stop II-8 of Bassett et al. (1974). Site 5 is the disused quarry in the Upper Llandeilo Flags 0.3 km north north-east of Dynevor Castle (SN 615229). Note that Sites 1-5 correspond to homologous sites in Bergström et al. (1987) base of the Llandeilian Stage can be identified based on graptolites in the Llandeilian reference sections. Based on the ranges of a few brachiopods and trilobites, Rushton et al. (2000, fig. 8.16) suggested with question that the base of the Llandeilian Stage corresponded to a level in the upper Pebbly Sands unit of the Ffairfach Group. In the absence of diagnostic graptolites, conodonts have been used to locate the approximate level of the top of the Llandeilian Stage, which Bergström et al. (1987) assumed to be somewhere in the middle of the Lower Llandeilo Flags. Added to the uncertain range of the Llandelian Stage in its reference section is the fact that both the name and precise range of the H. teretiusculus Zone have been the subject of discussion in recent years. Because the zone index ranges from well below the base of the H. teretiusculus Zone to some distance into the overlying N. gracilis Zone, the range of the zone index does not coincide with what has generally been taken to be the interval of the H. teretiusculus Zone. Because of this confusing situation, some recent authors have redefined the graptolite zone classification of this interval (e.g., Goldman et al. 2015;Chen et al. 2016) and proposed the use of the designation of the Jiangtzegraptus (formerly Dicellograptus) vagus Zone for the stratigraphical interval just below the base of the N. gracilis Zone.
The Ffairfach Grits and Llandeilo Limestone sequence, which has an estimated total thickness of more than 700 m, consists of a lithologically diverse succession of limestone, sandstone, grit, conglomerate, and shale deposited in intertidal to sublittoral open shelf environments. Virtually all our samples were collected before 1980 from limestone intervals and are the same as those dealt with by Bergström et al. (1987). The most productive samples are from interbeds and lenses of limestone in the Ashes and Lavas and Rhyolitic Conglomerates formations of the Ffairfach Group at the well-known railway cut at Ffairfach (Loc. 1 in Figure 1a-b). Some samples from the Golden Grove Group were collected from outcrops of the Llandeilo Flags at Golden Grove (Loc. 2 in Figure 1a) and Dynevor Park (for details, see Bergström et al. 1987). Important collection sites in this latter area are shown in Figure 1c  . For further information about localities and sample levels, see the Appendix. We have also had full access to the conodont collections from the Narberth area recorded by Bergström et al. (1987) as well as Bergström's samples from the Castell Limestone at its type locality near the quarry entrance at the northern end of Abereiddi Bay in Pembrokeshire (Nat. Grid Reference SM 795301; Loc. 6 in Figure 1d). For comparative purposes we have also investigated Bergström's (1971) Sandbian (early Caradoc) conodont collections from the Caradoc type area in the Onny Valley (Loc. 7 in Figure 1d) and at Evenwood Quarry (Loc. 8 in Figure 1d) 13 km south-east of Shrewsbury in South Shropshire (cf. Dean 1958) as well as numerous other collections, especially from Sweden.

Summary of the conodont biostratigraphy of the Ffairfach and Golden Grove groups and associated strata
As noted in Bergström et al. (1987), relatively numerous samples from calcareous beds in the Upper Ffairfach and the Golden Grove groups have yielded a fair number of biostratigraphically diagnostic conodonts although the frequency of conodonts in individual samples varies greatly from zero to relatively abundant. Whereas the conodont element yield from the relatively impure limestones in the Llandeilo Flags formations as a rule is not great, several of the limestone interbeds and calcareous lenses in the Upper Ffairfach Group produced rich and well-preserved conodont faunas. The elements exhibit a Colour Alteration Index (CAI; cf. Epstein et al. 1977) of about 5 (Bergström et al. 1987). The species identified and their frequency in the investigated samples are listed in Table 1.
The biostratigraphical evidence provided by the conodont collections studied herein has been discussed in some detail by Bergström et al. (1987) and their conclusions are still valid. Therefore, we here restrict our biostratigraphical comments to some major points.
The conodont species association in the samples from the Upper Fairfach Group includes the biostratigraphically important species Eoplacognathus lindstroemi and Baltoniodus prevariabilis that suggest correlation respectively with the Eoplacognathus lindstroemi Subzone of the Pygodus serra Zone and the Sagittodontina? kielcensis Subzone of the Pygodus anserinus Zone. In terms of standard graptolite zones, these subzones correspond to the upper part of the J. vagus Zone. The presence of the Subzone index species Amorphognathus inaequalis in the Flags and Grits Formation and in a sample from the Rhyolitic Conglomerates Formation in the uppermost Ffairfach Group, as well as in the Llandeilo Limestone at Golden Grove and at Dynevor Park, indicates the presence of the A. inaequalis Subzone of the Pygodus anserinus Zone. In Baltoscandia, the base of the Nemagraptus gracilis Zone, which coincides with the base of the Sandbian Global Stage, corresponds to a level in the lower to middle part of the A. inaequalis Subzone (Bergström 1983). In view of the fact that the Upper Llandeilo Flags Formation is overlain by the Dicranograptus Shales, which contain graptolites of the upper N. gracilis Zone but from which no conodonts have been recorded, the biostratigraphical evidence at hand indicates that virtually the entire Llandeilo Group corresponds to the N. gracilis Zone and is of Caradoc age in terms of the regional British series classification. The Llandeilian Stage, in its type area, would include only the very lowermost part of the Lower Llandeilo Flags Formation and possibly, a probably undeterminable part of the Ffairfach Group.
In an attempt to clarify the biostratigraphy around the base of the Caradoc Series, a level defined as the base of the N. gracilis Zone in South Wales, Bergström et al. (1987) reviewed the ranges of conodont taxa in samples from the calcareous succession in the Narberth area approximately 60 km west-southwest of Llandeilo. Although the faunas proved to have a relatively low diversity of conodonts, the presence of Amorphognathus tvaerensis, Baltoniodus variabilis, and Eoplacognathus elongatus in the upper portion of the Bryn-banc Limestone of the Narberth Group was interpreted by Bergström et al. (1987) as indicating the presence of strata corresponding to the Baltoniodus variabilis Subzone of the Amorphognathus tvaerensis Zone, hence an interval a little above the base of the N. gracilis Zone, and by implication, a little above the base of the Caradoc Series. This dating is consistent with the fact that the same interval of the Bryn-banc Limestone has yielded a few shelly fossils of early Caradoc affinity. However, more detailed investigations are needed to clarify if the successions in these quarries could serve as a standard reference for the base of the Caradoc Series. As is well known, in its typical development in the Welsh Borderland, the base of the Caradoc Series is at a major unconformity of regionally variable magnitude. Unfortunately, at Narberth, as well as elsewhere, the base of the Caradoc succession does not appear to coincide with a conodont zone or subzone boundary, being located biostratigraphically somewhat above the base of the Pygodus anserinus Zone. Our interpretation of the stratigraphy of the study interval in South Wales is illustrated in Figure 2.

Systematic Palaeontology
Most of the species listed in Table 1 are well known taxa that have already been adequately described in the conodont literature. Accordingly, there is no need to describe and comment on    the taxonomy of these taxa although we provide illustrations of representative specimens of most of these species, which include Baltoniodus prevariabilis (Fåhraeus, 1966) (Moskalenko, 1973) ( Figure 9N), and Venoistodus venustus (Stauffer, 1935) ( Figure 10M, O

Remarks on the early evolution of the genus Amorphognathus (Branson and Mehl, 1933)
Amorphognathus is a genus globally distributed and the short stratigraphical range of several of its species makes it one of the biostratigraphically most useful conodont genera in the Middle and Upper Ordovician. In the standard and widely used North Atlantic Upper Ordovician conodont zonation proposed by Bergström (1971), several zones were based on species of this genus. As shown by Bergström (1983, fig. 1), the oldest typical species of Amorphognathus, A. inaequalis, forms the beginning of an evolutionary lineage of a series of conodont species in which taxa grade into each other in stratigraphical order. These taxa include A. tvaerensis Bergström, 1962, A. superbus (Rhodes, 1955), and A. ordovicicus Branson and Mehl, 1933, and the evolution of these taxa in the Amorphognathus lineage is well documented (see Ferretti et al. 2014a;Bergström and Ferretti 2017). However, the evolutionary origin of this lineage remains enigmatic. Lindström (1977a, p. 22) interpreted Amorphognathus as derived from Baltoniodus but his concept of Amorphognathus was much wider than that favoured by conodont specialists today. The suggestion by Bergström (1983) that A. inaequalis might have evolved from Amorphognathus (now Sagittodontina?) kielcensis (Dzik, 1976) seems less likely now than 35 years ago but a true ancestor of Amorphognathus is not yet firmly established. For instance, the poorly known 'Polyplacognathus' angarense Moskalenko, 1984(cf. Sennikov et al. 2015 from the middle Darriwilian Mukteian horizon in Siberia has a P element that exhibits a superficial similarity to a species of Amorphognathus but its apparatus remains unknown. Our collections of topotype material of A. inaequalis and wellpreserved specimens of A. tvaerensis also include a few different ontogenetic stages. Study of these revealed that during the ontogeny, there were significant morphological changes both in the amorphognathiform (Pa) (Figures 3-4) and holodontiform (M) ( Figure 5) elements. When dealing with other faunas, recognition and illustration of such growth stages may assist in making correct identification of particularly more or less fragmentary elements of different growth stages of these taxa. Genus Amorphognathus Branson and Mehl, 1933 Type species. Amorphognathus ordovicicus Branson and Mehl, 1933. Remarks. For a description and discussion of this species based on new collections from its type locality, see Leslie and Bergström (2005). Rhodes, 1953 ( Figures  Material. 118 specimens. Description. For a concise diagnosis and description of this species, see Lindström (1977a). Both blade and non blade amorphognathiform (Pa) elements are present in our investigated collections. Our material includes elements that closely match the single Pa blade element ( Figure 3H) that was selected by Rhodes (1953) as the holotype of the species. An almost complete Pa blade element (Figures 3G and 9C-D) has a slender blade-like anterior process and a longer platform-like posterior process, which has proximally a distinct, but low, cusp. The posterior process has a single row of denticles, which runs in the middle of the platform from the cusp to the end of the element. From the cusp diverges a short bi-lobed postero-lateral process that is less than half as long as the posterior process. The bifid postero-lateral process has a posterior lobe with a single row of denticles running almost parallel to the posterior process denticulation. An incipient anterior lobe is developed as well, with a central ridge (our material, Figures 3G and 9C-D and the specimen illustrated in Bergström and Orchard 1985, pl. 2.2:14, Figure 3I and in Bergström et al. 1987, pl. 18.1:9) or a denticle (holotype, Rhodes 1953, pl. 22:204, Figure 3H). The amorphognathiform element illustrated by Lindström et al. (1974, pl. 2:1) lacks this lobe, representing probably a less evolute form ( Figure  3J). All specimens document an incipient foreword bulge in the posterior process on the opposite side of the postero-lateral process. The antero-lateral process is not completely preserved but appears diverging. Otherwise, the upper surface of the entire Pa element is smooth.

Amorphognathus inaequalis
Homologous elements of A. tvaerensis possess a double-branched postero-lateral process, having all branches distinctly denticulated. The anterior lobe of the postero-lateral process is of equal size of the posterior lobe (our material, Figures 3E-F and 11B, E) or longer (e.g., Bednarczyk 1971, pl. 6:6a, Figure 3B Figure 3A). The postero-lateral process appears to migrate from a posterior position in A. inaequalis to a more anterior position in A. tvaerensis, in the latter developing at about mid-length of the element main axis.
Non blade Pa elements of A. inaequalis are preserved in our material with incomplete specimens documenting an anterior and posterior process of similar size and with a straight alignment (Figures 4I-J and 9A-B). The posterior process possesses a single incipient adenticulated bulge, located at mid-length on the outer side, diverging with a step-like 90° angle from the posterior process. The same feature is documented also in the material described by Lindström et al. (1974, pl. 1, fig. 9) ( Figure 4K). This bulge will give rise in A. tvaerensis to the diagnostic second posterior process ( Figure 4E), possibly through intermediate forms where this process is present but not denticulated ( Figures 4F and 11C). An additional lobe is common in A. tvaerensis on the anterior side of the second posterior process ( Figures 4A, C-H and 11C, J).
It should be noted that the non blade Pa element is not present in the Rhodes (1953) collection but it is represented in the material described by Lindström et al. (1974).
The unpaired Pb elements are relatively abundant in our collections. Sinistral Pb elements ( Figure 9E-F) have a more rounded basal margin compared with the straighter basal margin of the dextral elements (cf. Figure 9G-I). The same difference has been illustrated in Pb elements of other species of Amorphognathus (e.g., Bergström 1962;Serpagli 1991, 1999;Bagnoli et al. 1998;Ferretti et al. 2014aFerretti et al. , 2014bFerretti et al. , 2014cBergström and Ferretti 2015). The Pb elements have a short, denticulated lateral process whose edge extends up to the cusp ( Figure  9H) sometimes revealing a small rudimental expansion located on the outer edge of the cusp ( Figure 9E, G). The anterior surface of both the anterior and posterior processes has a well-developed ledge.
A single holodontiform (M) element was recovered in a sample from the Ffairfach Group (Figures 5J, 9J-K). It has three processes with many subequal-sized denticles and a cusp of similar size as the denticles. The long anterior process is uniformly denticulated along its entire length with small, fused, and upwards-directed denticles. In outer-lateral view, the upper denticles are oriented posteriorly and the whole element is strongly arched ( Figure 9K). There is also a rather long and denticulated posterior process in the M element. Lindström et al. (1974) remarked that the holodontiform elements of A. inaequalis have a more equant denticulation and a smaller cusp than those of A. tvaerensis. However, the specimen of A. inaequalis from the French material is photographed in lateral view (pl. 2:7, Figure 5L), and therefore hardly informative to illustrate denticle size. The specimen illustrated by Bergström et al. (1987, pl. 18.1:10) ( Figure 5K) shows an anterior process lacking a distinct denticulation, but denticles on the oral side appear to have similar size.
There are only a few S elements ( Figure 9L-M) in our Llandeilo collection. These clearly correspond to the homologous elements in other species of Amorphognathus. We have nothing to add to the descriptions of these elements published by previous authors.
Remarks. Compared with other species of Amorphognathus, A. inaequalis has been rather poorly known. This is undoubtedly due to the fact that only a comparatively limited number of elements of this species has been recorded. The holotype of the species, collected from the Llandeilo Limestone at Golden Grove (Site 2 in Figure 1), is a blade Pa element (Rhodes 1953, pl. 22: 204; Figure  3H) that is almost complete, possibly missing only a minor part of the antero-lateral process. Rhodes (1953) also described a single Pb element from the same locality that undoubtedly belongs to the same species. Lindström et al. (1974) were the first to propose a multielement reconstruction of the apparatus of A. inaequalis based on collections from the Postolonnec Formation of the Armorican Massif in Brittany and they described and illustrated all element types in the apparatus. We agree that their specimens, although they come from quite a different region than that of the holotype, represent A. inaequalis.
Bergström (1964,1971), Bergström and Orchard (1985) and Bergström et al. (1987) recorded the species from several localities in South Wales. Lindström (1977a) gave a succinct description of the species and discussed its relations to A. tvaerensis. Dzik (1984) regarded the Welsh and French specimens as belonging to Rhodesognathus Bergström and Sweet, 1966. However, his few and very fragmentary specimens come from the B. variabilis Subzone of the Môjcza Limestone in Poland, hence from a younger interval than the type stratum in Wales, and the Polish species record clearly needs to be confirmed by better specimens. The Pb elements of A. inaequalis documented in this paper show a rudimental expansion on the lateral surface of the cusp, that cannot be confused with the separate denticle located on the anterior process of Rhodesognathus to which the lateral process connects in that genus. The lateral process of A. inaequalis is clearly Note that the precise relations between the graptolite and conodont biostratigraphy are still not precisely known in the Welsh study successions due to the lack of records of biostratigraphically important graptolites in these successions. Also note that the stage classification of the study successions follows Williams et al. (1972).
connected directly with the cusp ( Figure 9H). In addition, the M element does not present the diagnostic reclined cusp. However, we cannot exclude that A. inaequalis might have represented the ancestral form of both lineages. Specimens of A. inaequalis have also been recorded, but as far as we know not illustrated, from an interval just below the appearance level of A. tvaerensis in Estonia (Männik 2003;Männik and Viira 2005;Viira 2008) and even an A. inaequalis Zone between the Pygodus anserinus Zone and the A. tvaerensis Zone has been introduced in the Estonian conodont zone succession (Viira 2008).
The species has not yet been firmly identified in Sweden although it has been used as a designation for an A. inaequalis Subzone in the upper part of the P. anserinus Zone in the standard conodont succession in central Sweden (cf. Bergström 2007).
Description. This species was introduced by Bergström (1962) based on specimens from the Ludibundus (now Dalby) Limestone at a locality in southeastern Sweden. According to the original description, the non blade-type amorphognathiform (Pa) element consists of five processes: one anterior, one posterior, and three lateral processes. Among the latter, one bilobate antero-lateral process branches from the anterior process, another bilobate process branches from the anterior part of the posterior process (creating the '1 st posterior lateral process') on the opposite side of the antero-lateral process, and one unilobate postero-lateral process ('2 nd posterior lateral process') diverges from a middle point of the posterior process on the same side of the bifid postero-lateral process. All lateral processes are shorter than the anterior and posterior processes and carry a single row of denticles ( Figures  4A-E, G-H, 11J)  F' Bergström 1971 HOLOTYPE Figure 6. Outline drawings of elements of Baltoniodus alobatus (Bergström, 1971). ridge ( Figures 4F and 11C). Importantly, in this type of Pa element, there is a single uni-lobe postero-lateral process, not a double one as in A. inaequalis. The blade-type Pa element is similar to the non blade-type Pa element except it lacks the uni-lobe postero-lateral process. Apparently almost complete blade Pa elements of different growth stages occur in our study collections. In blade Pa elements, the anterior process is shorter than the posterior one and is strongly laterally compressed. The antero-lateral process, which branches from the anterior process, is as a rule broken off in the Welsh specimens but in a single immature specimen (Figures 3E and 11B) it appears to be unilobate. The posterior process may be twice as long as the anterior one and carries a central row of nod-like denticles that continue to the end of the posterior process. In some specimens, the surface of the posterior process opposite and posterior to the postero-lateral process is widened ( Figures 3E-F and 11E). The postero-lateral process is bifid with single central rows of denticles on each lobe. A prominent ledge runs along the margin of the processes.
The unpaired Pb elements resemble those in A. inaequalis.
The holodontiform (M) elements are represented by only a few specimens ( Figure 5G-I). which have three denticulated processes. The anterior process is long and straight and carries a row of basally fused denticles which are continuous with the apical denticulation. The upper margin of the M element has a row of denticles of somewhat large size which is flexed postero-laterally. Among these denticles, the cusp can be distinguished although it is not very prominent. Some specimens from Abereiddi Bay have a reclined marginal apical denticle ( Figure 5I) as was noted by Bergström et al. (1987).
Remarks. The relations to A. inaequalis, the apparent ancestor of A. tvaerensis, have been discussed above. Leslie (2000) noted that the M element of A. tvaerensis has at least two relatively large denticles adjacent to the cusp whereas A. superbus has a single such denticle and A. ordovicicus lacks such denticles adjacent to the cusp.
Based on collections from China and Sweden, Zhang (1998) gave Eoplacognathus a more restricted scope in recognising two species previously referred to Eoplacognathus (E. reclinatus (Fåhraeus, 1966) and E. robustus Bergström, 1971) as representing a separate lineage, for which she proposed the generic designation Baltoplacognathus. For another lineage, which includes mainly Chinese taxa and which was recognised as a side branch of Eoplacognathus by Bergström (1983, fig. 2), she introduced the designation Yangtzeplacognathus. The species E. lindstroemi, and E. elongatus remained in Eoplacognathus following Zhang's (1998) classification, which has been adopted by some, but not all, subsequent authors. Because the species present in our Welsh collections, E. lindstroemi, is the type species of Eoplacognathus, its generic classification is obvious whether or not Zhang's (1998) revised classification is accepted.
In some respects, Eoplacognathus is reminiscent of Amorphognathus Branson & Mehl, 1933, but, as noted by Lindström (1977b, the main distinguishing features from Amorphognathus are the great thickening of the platform ledges, the much narrower basal cavity, the conspicuous development of a laterally pointing lobe in the Pa element, and the presence of a very long anterior process in the Pb element. For many years, Eoplacognathus was interpreted to lack ramiform elements in the apparatus (e.g., Bergström 1983). However, in recent years, some authors have proposed that the apparatus of at least some species of this genus included also ramiform and geniculate elements. For instance, Dzik (1994) reconstructed the apparatus of the species E. zgierzensis Dzik, 1976 to include, apart from Pa and Pb elements, an M element with three denticulated processes, and a set of four types of S elements. Stouge and Bagnoli (1990) identified S and M elements in Lenodus (herein Eoplacognathus) pseudoplanus. Viira et al. (2001), in a study of conodonts from the Middle Ordovician Mäekalda section in Estonia, illustrated the apparatus of Eoplacognathus pseudoplanus as having ramiform elements (referred to as Sb, Sc1, and Sc2 elements) and a geniculate M element.
In another study, based on a large conodont collection from south-central China and Sweden, Löfgren and Zhang (2003) recognised seven morphologically distinct element types, and possibly a total of 17 elements, in the apparatuses of Lenodus variabilis, Yangtzeplacognathus crassus and Eoplacognathus pseudoplanus. Separation of the non-platform elements of these genera from the homologous ones in the co-occurring Baltoniodus specimens was possible based on the presence in the former genera of wart-like extensions at the top of the cusp (cf. Löfgren 1990). Heredia et al. (2014) reported the presence of S and M elements in E. suecicus, E. robustus, and E. lindstroemi and Heredia and Mestre (2019) described full apparatuses of E. robustus and E. lindstroemi.
Several authors (e.g., Dzik 1994;Löfgren and Zhang 2003;Heredia and Mestre 2019) have discussed the suprageneric classification of these Darriwilian platform genera but at the present time, there is little agreement. The species E. lindstroemi clearly evolves into E. elongatus and several authors (e.g., Barnes and Fåhraeus 1975) have suggested that the latter species is the ancestor of Polyplacognathus ramosus Stauffer, 1935 although transitional forms between the two latter taxa have not yet been recorded.
The latter species has generally been interpreted to have a bimembrate apparatus but its platform elements are commonly associated with ramiform elements originally described as Trichonodella? tricurva Schopf, 1966 andTetraprioniodus breviconus Webers, 1966. Both these taxa look like S elements and they were later referred to a new genus, Schopfodus Leslie, 1996. Although Leslie (1996 expressed the opinion that these elements might not be a part of the apparatus of P. ramosus, we feel that at the current state of knowledge, this possibility cannot be dismissed. If this idea proves to be correct, the S elements of this taxon differ conspicuously from those assigned to early species of Eoplacognathus. This obviously would support the current generic separation of these two genera. Eoplacognathus lindstroemi (Hamar, 1964) (Figures 7A-V, 8A -L) Mestre: fig. 4. Description. The Ffairfach conodont collection contains a surprisingly large number of well-preserved elements of this geographically widespread species. Many of the elements are complete and the presence of growth stages illustrates the ontogeny of both the Pa and Pb elements. Both the Pa and Pb elements of this species are easily recognisable also in fragments by their thick and robust platform ledges and the narrow, slit-like basal cavity. Even the smaller specimens revealed to have yet the diagnostic features well delineated, both in Pa and Pb elements.
Because elements of this species have been described repeatedly in the literature, we restrict ourselves to offer only some specific comments. The Pa elements have a straight antero-posterior axis and a welldeveloped postero-lateral process which is almost as long as the posterior process. The posterior process is expanded laterally ( Figure 7B, D) with a bifurcating terminal denticulation in many mature specimens ( Figure 7E). According to Lindström (1977b) the Pb elements have posterior and lateral processes of approximately equal length, diverging from one another at an angle 'Y-shaped.' In our collection, this applies only to the dextral Pb elements, which have two processes of similar size that represent the upper 'arms' of the Y. Sinistral elements have lateral processes that are always shorter than the posterior one. The posterior process of sinistral elements lacks lateral expansion of the posterior part of its distal corner also in mature specimens. In both sinistral and dextral elements, the antero-posterior axis is lightly sinuous.
Despite the large number of platform elements in our collections, there are only a few M and S elements that may possibly be a part of the E. lindstroemi apparatus based on previous descriptions of such elements. Hence, a single broken M element recovered in sample 66-2 has a distinct anterior margin denticulation, a proclined cusp and a typical continuity in the curve of the anterior process and cusp that serve to separate this element from the M element in Baltoniodus.
No Sa element was recovered but a single Sb element ( Figure 8V) from the same sample has three denticulated processes, which are connected by a basal sheet. These elements are distinguished from the corresponding element in Baltoniodus by the presence of a denticulated lateral process. A possible Sc element ( Figure 8W) was recovered from the same sample.
A polygonal micropattern has been observed on the platform ledges of the Pa and Pb elements but not in the M and S elements. It should be noted that Löfgren and Zhang (2003) observed such micro-ornamentation in S elements of Lenodus antivariabilis. There are also thin parallel striae running from the platform ledges to and along the denticles in both the Pa and Pb elements.

Remarks.
In an attempt to better define the differences between the M and S elements of E. lindstroemi and corresponding elements of Baltoniodus, which are superficially closely similar, we examined large collections of Baltoniodus from the Dalby Limestone at Fjäcka, in south-central Sweden (Bergström 2007). In the latter genus, the Pa and Pb elements are quite different from the homologous elements in E. lindstroemi. In Baltoniodus, the M elements have a long, generally adenticulated, anterior process ( Figure 6C-E, H-M) that in some elements may have a distal crenulation. These elements have a flat inner face and a strongly convex outer face due to the presence of a conspicuous lateral bulge on the base. In inner view, both the inner and outer basal margin are visible. In some specimens, the cusp is twisted with respect to the base.
The Sa elements of Baltoniodus are symmetrically alate ( Figure 6N-P) with three denticulated processes. In many specimens, the posterior process has a hindeodellid type of denticulation.
The Sb elements (corresponding to Sc2 elements of Löfgren; Figure 6Q-V) are asymmetrical, strongly compressed laterally, and have a more or less undenticulated lateral rib near the anterior margin.
The Sc elements (corresponding to the Sc1 elements of Löfgren; Figure 6W-B') are similar to the Sb elements but differ from the Sb elements in lacking the outer lateral rib. In some cases it is difficult to separate Sb and Sc elements.
Sd elements (corresponding to Sb elements of Löfgren; Fig. 6C'-F') are tetracostate, having four denticulated processes that may be symmetrically or asymmetrically arranged in relation to the posterior process. In this type of element, the posterior process in some elements has a hindeodellid type of denticulation.
Genus Erismodus Branson and Mehl, 1933 Type species. Erismodus typus Branson and Mehl, 1933. Erismodus cf. E. bergstroemi (Savage and Bassett, 1985) ( Figure 10P- Material. 141 elements. Remarks. One of the more notable taxa in our collections from the Llandeilo region includes frequently very large but uncommon elements that are currently difficult to classify. Bergström (1964) first described these elements and interpreted them as representatives of the North American genus Erismodus based on the similarity of some of these elements to E. typus, the type species of Erismodus. Bergström (1964) also illustrated another type of element that he referred to another North American genus, namely Chirognathus Branson & Mehl, 1933. We now believe that these types of elements are likely to belong to the same multielement apparatus but the currently available Welsh collections are insufficient to reconstruct the composition of this apparatus. Added to this problem is the fact that the apparatus of E. typus is not well known. Interestingly, Savage and Bassett (1985) described very similar conodonts from the Caradocian of South Shropshire which they named Plectodina bergstroemi. Our collections from the lower Caradocian of that region includes some representatives of this taxon but the collections at hand are insufficient to determine if the Caradocian species is the same as the Llandeilo one. However, we believe that these distinctive conodonts are not representatives of Plectodina but are more likely to belong to Erismodus. Pending the availability of more informative collections, we prefer to tentatively classify the Llandeilo specimens as Erismodus cf.
Remarks. In Bergström's (1964) collections from the Castell Limestone at Abereiddi (formerly Abereiddy) Bay, there are four S elements that closely resemble those of the North American species P. inflexus Stauffer, 1935. It should also be noted that our Castell Limestone elements exhibit some similarity to the tr (S) elements of Phragmodus polonicus Dzik, 1978(cf. Dzik 1994 fig. 25b-c). To confirm the identification, additional collecting is needed to recover the other elements of the apparatus. Because this is the only potential record of this species in Europe, it has considerable biogeographical interest. This species has not been recovered in the Llandeilo region.

Material. 2267 elements.
Description. This species is the most abundant taxon in our Llandeilo region collections and all components of the apparatus are present. The blade-like Pa elements ( Figure 9Q-R) have a somewhat expanded basal cavity that extends, especially in mature specimens, along the under side of the anterior and posterior processes. These processes are of about equal length and form an angle of 130-140 degrees with each other in lateral view. In most cases, the anterior process is slightly wider (higher) than the posterior one. Both processes carry about half a dozen short and erect denticles that tend to be less densely spaced in the posterior process. The Pb elements in our collection are broken but appear to be of dichognathiform type ( Figure 9S-T). The M element has a well-developed, more or less straight posterior process with about half a dozen very strongly reclined, densely spaced, denticles of about half the length of the cusp ( Figure 9V, B'). One face of the base of the cusp is expanded laterally to form a bulge, the pointed basal corner of which forms an angle of slightly less than 50 degrees with the anterior margin of the cusp. The S elements form a transition series of four types of elements. The Sa element ( Figure 9X) is alate but lacking posterior process. Its two lateral processes have a small number of stout and straight denticles of about half the length of the cusp. It should be noted that the specimen illustrated in Figure 9X is essentially identical in morphology to Rhodes' holotype of Trichonodella flexa (cf. Bergström 1964, text-fig. 20). The Sb element ( Figure 9Y-Z) is conspicuously digyrate with two asymmetrically located lateral processes and no posterior process. The Sc element ( Figure 9U-W) is dolabrate with a series of slightly reclined denticles along the posterior process. There is no distinct Sd element but slightly asymmetrical elements that are morphologically intermediate between the Sb and Sa elements ( Figure 9A') are present.
Remarks. In general organisation, the apparatus of P. flexa agrees with that of several species of Plectodina best known in the North American Midcontinent faunas, such as P. tenuis Branson and Mehl, 1933. Components of the apparatus of these taxa have been described in a multitude of form taxonomic names (cf. Bergström and Sweet 1966) but the multielement taxonomy of species of this genus has not yet been fully established. Especially in the appearance of the Pa element, P. flexa is clearly different from, for instance, P. tenuis, P. aculeata Stauffer, P. florida Sweet, P. aculeatoides Sweet and P. bullhillensis Savage and Bassett but in some characteristics, it is similar to specimens from the Welsh Borderland identified as P. tenuis by Savage and Bassett (1985). Pending a needed revision of the genus, we herein use the species name introduced by Rhodes (1953) for the late Darriwilian-early Sandbian species from the Llandeilo region.

Remarks.
A complete set of elements of the Plectodina apparatus was documented in the material from Onny Valley and a single element was recovered from the Evenwood Quarry. Pa elements do not reveal the expanded basal cavity that typifies P. flexa (compare Figures 9Q-R and 12J-K). Furthermore, Pb elements are typical arched dichognathiform elements that preserve a denticulate anterior and posterior processes but are lacking the plica observed in the corresponding element of P. flexa (compare Figures 9S-T and 12L-M). On the contrary, a ridge is visible in outer-lateral view. M and S elements are indistinguishable from the homologous elements of P. flexa. This species from Onny Valley might be new, but it is insufficiently known at present.

Conclusions
The principal results of the present study may be summarised as follows: 1. The examination of all available conodont collections from the reference sections of the British Llandeilian Stage confirms the biostratigraphical conclusions presented by Bergström et al. (1987), namely that in its classical development in the Llandeilo region, the age of the Llandeilo Flags corresponds to an interval ranging from the Sagittodontina? kielcensis Subzone of the Pygodus anserinus Zone to the Baltoniodus variabilis Subzone of the A. tvaerensis Zone. The underlying Ffairfach Group yields diverse conodont faunas of the Eoplacognathus lindstroemi Subzone of the Pygodus serra Zone and its upper part is of late Darriwilian age. Following the common practice of defining the base of the regional Caradoc Series as the base of the Nemagraptus gracilis Zone, most of the type Llandeilian succession corresponds to the lower Caradoc, only the lowermost part of the Llandeilan being of pre-Caradoc age.
2. The conodont faunas of the upper Ffairfach Group and the overlying Llandeilo Flags of the Golden Grove Group include about 20 multielement species. Apart from the presence of a few widespread coniform species, these faunas are characterised by the presence of stratigraphically important platform taxa of Amorphognathus, Eoplacognathus, and Icriodella along with representatives of the widespread genus Baltoniodus. Because these genera exhibited rapid evolution and the development of short-ranging species, they are very useful for precise biostratigraphical dating, especially as these taxa commonly are quite widespread geographically.
3. The investigated conodont faunas from Wales have their own provincial character. The biostratigraphically broadly equivalent Sandbian conodont fauna recently described from the Garn Formation of Anglesey (Bergström and Ferretti 2018) includes representatives of Amorphognathus, Eoplacognathus, and Baltoniodus but the presence of Periodon, Protopanderodus, Pygodus, and Spinodus gives the Garn conodont fauna a rather different appearance than those from the Llandeilo area. In view of its presumed geographical location off the microcontinent Avalonia between Baltica and Laurentia during Sandbian-early Katian time ( Figure 13) it is not surprising that the Avalonian conodont faunas dealt with herein share quite a few species with coeval Baltoscandic faunas (cf. Männik and Viira 2005;Bergström 2007) but most of the latter differ by the absence of, for instance, representatives of Plectodina and the common occurrence of Protopanderodus, Periodon and Pygodus. Equivalent faunas from the eastern Appalachian thrust belts in North America (e.g., Sweet and Bergström 1962;Bergström 1971) share quite a few species with the Welsh faunas dealt with herein but they also contain a variety of different species. On the other hand, equivalent faunas from the North American Midcontinent, for instance the Tulip Creek Formation of Oklahoma (Bauer 1987), have very little in common with the Darriwilian-Sandbian faunas of Wales. For further discussion of the global conodont biogeography of this stratigraphical interval, see Bergström (1973Bergström ( , 1983, Dzik (1993), and Sweet and Bergström (1974).