NEW ACROPOMATIFORM FOSSILS FROM THE UPPER KUEICHULIN FORMATION (LOWER PLIOCENE), NORTHERN TAIWAN

To cite this article: Přikryl T., Lin C.-H., Hsu C.-H. & Lee S.-W. (2024) - New acropomatiform fossils from the Upper Kueichulin Formation (Lower Pliocene), northern Taiwan. Riv. It. Paleontol. Strat. , 130(2): 211-229.


IntroductIon
In contrast to the extensive diversity observed in the contemporary fish fauna, the records of fish fossils in the tropical-subtropical West Pacific region are notably limited and elusive.Despite ongoing efforts and the accumulation of knowledge related to some taxa, such as elasmobranchs (e.g., Marramà et al. 2018;Kocsis et al. 2019;Lin et al. 2022b), the overall understanding of palaeoichthyology in this region remains conspicuously sparse.As a result, significant portions of this vast expanse continue to be terra incognita in the field of palaeoichthyology.
The island of Taiwan is situated within the Sino-Japanese Province (Toonen et al. 2016), a position shaped by a complex convergent boundary.The fossil fish remains found on the island are represented by teeth, otoliths, isolated bones, and very rarely, by articulated and semi-articulated skeletons (Lin et al. 2021(Lin et al. , 2022a;;Lin & Chien 2022).Noteworthy among these finds are fish skeletons from the Lower Pliocene deposits of the Kueichulin Formation in northern Taiwan, as mentioned in Lin et al. (2021: fig. 3A).Specifically, this study focuses on the investigation of five individuals discovered from this formation, aiming to elucidate their significance within the broader context of palaeoichthyofauna.
The extant members of the order Acropomatiformes exhibit a widespread distribution across both shallow and deeper waters in tropical and subtropical regions worldwide.Several species within this order possess light-emitting organs on their ventral sides (Okamoto & Gon 2018;Ghedotti et al. 2018;Smith et al. 2022).Notably, the formal recognition of the order Acropomatiformes is a relatively recent development, primarily established based on molecular analysis (see Smith et al. 2022 for details).Consequently, the composition and relationships among the families within this order have become a focal point of discussion (e.g., Davis et al. 2016;Mirande 2016;Ghedotti et al. 2018;Smith et al. 2022).Based on the latest findings, the order is now recognised to comprise 20 families (Smith et al. 2022).Furthermore, the ongoing taxonomic exploration has led to the continuous discovery of new species within this extensive group (e.g., Yamanoue 2016;Schwarzhans & Prokofiev 2017;Mediodia et al. 2024).
For the fossil record of the order Acropomatiformes, pertinent otolith evidence appears to be widely distributed, with early putative acropomatid otoliths such as "Acropomatida" bagassiana Nolf, 2003 from the Santonian of Spain (Nolf 2003), and Plesiopoma otiosa Schwarzhans, 2010 from the Maastrichtian Bavaria, Germany (Schwarzhans 2010).However, it is not until the Late Palaeogene that the otoliths exhibit a closer affinity to those of their extant counterparts.In the West Pacific, acropomatiform otoliths are predominantly represented by the genus Parascombrops (see Schwarzhans & Prokofiev 2017), including Parascombrops schwarzhansi van Hinsbergh & Helwerda, 2019 from the northwest Philippines (van Hinsbergh & Helwerda 2019) and Parascombrops ohei Schwarzhans & Prokofiev, 2017 from eastern Japan (Mitsui et al. 2021).Otoliths of Acropoma are also commonly found in the Lower Pleistocene deposits in southwestern Taiwan (Lin et al. 2018).More recently, Lin et al. (2023) reported otoliths of Malakichthys, Synagrops, and Acropo-ma from a Lower Pliocene deep-sea assemblage in southern Taiwan.
Nevertheless, the skeletal fossil record of Acropomatiformes is not significant worldwide, with only a handful of records primarily originating from families Acropomatidae and Stereolepididae (for details, see Discussion).The studied specimens represent a noteworthy addition to the fossil record of the Stereolepididae.Consequently, the main purpose of this paper is to document, discuss, and evaluate the acropomatiform specimens discovered from the Tachi site (Kueichulin Formation; northern Taiwan).

GeoloGIcal settInG
The sampling site (Tachi) is situated within the Upper Kueichulin Formation (locally also recognized as the Erchiu Formation), which is Early Pliocene in age (Fig. 1).The formation is widely exposed along the upstream of the Dahan River and it primarily consists of intensely bioturbated muddy sandstone with thick-bedded mudstone and interbedded layers of sandstone and mudstone (Pan et al. 2015;Chen 2016).In addition to abundant trace fossils and shell fragments (Wang 1983(Wang 1984b;;Chen 1986), articulated marine fossils, exceptionally well-preserved within nodules, are also commonly found in the strata (Chen 2016).
These fossiliferous nodules often are encapsulating in situ marine fossils, predominantly crustaceans (Hu & Tao 1996).The fish skeletons described in the present paper are from such nodules.The locality also yields other marine fossils not being encased in nodules, such as echinoids (Wang 1984a(Wang , 1985;;Hsu et al. in review), mollusks (Hu 1992), cuttlebones (Lee unpublished data), and fish otoliths (Lin unpublished data).In summary, based on both sedimentology and ichnology, the sedimentary environment of this locality indicates relatively shallow water depths, positioned above the shoreface-offshore transition zone (Chen 1986;Pan et al. 2015;Chen 2016).

MaterIal and Methods
The fossils were discovered by local amateur fossil collectors in the 2010s in Tachi, Taoyuan, northern Taiwan.These specimens collectively represent five individuals, with four encased within two nodules and one isolated skull.These fossils are deposited in the Na-tional Museum of Marine Science and Technology, Keelung, Taiwan (under the registration code NMMST) and the Biodiversity Research Museum, Academia Sinica, Taiwan (ASIZF).
The first nodule, designated as NMMST 20190-3001 (Fig. 2), preserves the semi-articulated remains of a single fish with a measurable length of 101 mm.The head length is approximately 25 mm, with the neurocranium measuring around 23 mm in length.The second nodule, NMMST 20190-3005 (Fig. 3A, B), contains the remains of three individuals, especially partially preserved skulls.Two neurocrania are more or less complete, with a complete neurocranium measuring about 30 mm in length, along with fragments of the postcranial skeletons.Besides these two nodules, there is yet another partially preserved skull (ASIZF0100722, Fig. 3C).Despite the fragmented nature of the specimens, all individuals are considered to be monospecific due to morphological similarities of the available parts of the skulls.The dislocation of scales, together with disarticulation of vertebral column, abrasion of some bones, and partial loss of the fin spines and rays, suggest potential post-mortal transport and/or alternations to the original positioning of the body.Therefore, it is reasonable to expect that the body proportions and certain meristic characters might have been different in their original state than is currently observed.
Paratypes: NMMST 20190-3005 preserving remains of three individuals and ASIZF0100722 preserving a single individual.All specimens are from the type locality.
Etymology: The species name refers to "secret" in Latin, reflecting incomplete and mysterious nature of these fossils.

Diagnosis:
The new species differs from other species of the genus by having a significantly shorter ethmoidal area of the skull (less than 50% of the orbital area), strongly developed sphenotics with their depth in lateral view representing approximately half of the ethmoidal depth, and the strongly developed "pit and ridges" ornamentation in most of the dorsal surface of the frontal bones.Description.The fossils under examination contain remains of small fish with partially preserved skeletons (Figs 2-4).The estimated SL (based on the measurable length of NMMST 20190-3001) is about 110 mm.
The neurocranium.The neurocrania (Figs 5-7) exhibit dorsal surface dermal ornamentation characterised by numerous ridges and pits.This distinctive ornamentation is predominantly observed on the frontals, extending partly to other elements, such as parietals, sphenotics, parts of pterotics, and supraoccipital.The frontals are about one and a half times longer than their width, and their maximal width occur just in the postorbital section, precisely at the level of articulation with the sphenotic.The anterior margin of the frontals is convex, with a parabolic shape.Positioned anteriorly to the frontals is a recognizable short, but deep triangle-shaped mesethmoid, accompanied by laterally articulated lateral ethmoids, showing a concave anterior surface of the bone.The ethmoid area is short (representing maximally half the length of the orbit) and relatively deep (approximately the same as the orbit length).The frontals articulate posteriorly with the supraoccipital, parietals and pterotics.Posteriorly to the frontals, in the mid-line of the neurocranium, a clearly recognizable supraoccipital is present.The antero-dorsal part of the supraoccipital bears typical dermal structures, while the posterior half is thin and forms the dorsal margin of a short and very low supraoccipital crest.Laterally to anterior section of the supraoccipital are rhomboid-shaped ported by robust skeletal tissue (i.e.creating "riblike" thorns).Similarly, the anterior margin of the opercle is enforced by thickened bonny tissue-the opercle is therefore all together reinforced by three axes, which are located on the anterior edge and in the axes of the spines.The upper spine is dorsally completed by bony lamina with a convex dorsal margin.Notably, one of the opercles preserved in NMMST20190-3005 has undergone taphonomic reduction, retaining only the most ossified parts of the bone (see Fig. 8B).
The preopercle is sickle-shaped, with the vertical ramus much longer than its horizontal part (about four times as long).While spinulation and serration are not macroscopically visible, micro-CT examination (Fig. 5) reveals these structures along the posterior and ventral margins of the bone.The posterior section in the dorsal part of the preopercle is relatively finely serrated and spines become slightly larger in the corner of the bone and along the ventral margin of the horizontal limb (Fig. 5A, B).Posteriorly to the corner of preopercle, there is an articulated interopercle, but the details of its morphology is not clear.A rod-like element preserved in the specimen NMMST20190-3001, showing an antero-dorsally expanded part, is interpreted as a partially preserved hyomandibula, but this bone is incompletely preserved.Other parts of the skull are not determinable due to insufficient preservation.
The axial skeleton.The vertebral column is not completely preserved in any of the available specimens.Disarticulated remains preserved in NMMST 20190-3001 allow for the recognition of 8 abdominal and 10 caudal vertebral centrae, but the actual number of vertebrae was unequivocally higher.The vertebral centrae do not bear recognizable growth rings.The fossil also preserves the remains of a single predorsal bone, although the postcranial section is not sufficiently preserved, it is reasonable to expect other predorsals were presented.Approximately seven or eight moderately-developed large ribs are discernible.The caudal skeleton and caudal fin are not preserved.
The dorsal fin.The dorsal fin exhibits remains of ten spines, with the preservation suggesting that this count is either original or only slightly higher.Additionally, about 6 rays are recognizable.The base of the spinous part of the dorsal fin is much longer than its soft part.The articulation of the last spinous dorsal element is in the level just in front of the start of the anal fin.Length proportions of individual spines and rays are not determinable.
The anal fin.The anal fin is characterized by a short base.The fin itself is developed in the posterior-most third of the body, just below the soft part of the dorsal fin.It consists of three spines and seven or eight rays.The first spine is shortest than the second and third ones (see Table 1 for details of proportions).
The pectoral girdle and fin.The cleithrum is stick-like, more or less straight, dorsally enlarged to plate-like.The scapula is trapezoid, with a large central foramen.The supracleithrum widens postero-ventrally, with a length more or less the same as the maximal antero-posterior length of the opercle.Posteriorly to the right pterotic of the specimen NMMST 20190-3001, a plate lake posttemporal is recognizable.This posttemporal has two anteriorly oriented rami.The dorsal posttemporal ramus continues towards the epioccipital, although the articu-  lation is covered by sediment in all specimens.The ventral ramus extends towards the lateral extension of the pterotic.The pectoral fin is preserved only as fragments of its base, with recognizable remains of 16 or 17 rays.The shape of the fin itself is not determinable.
The pelvic fin and girdle.The position of the pelvic fin is just below the pectoral girdle.The basal plate has the shape of a strongly elongated triangle, with a well-developed post-pelvic process.The pelvic fins consist of one spine and five elongated rays, with their posterior-most tips reaching approximately half of the body length.
Scales.The body is covered by very small cycloid scales, with an approximate diameter ranging from 2 to 3 mm.These scales are preserved in larger areas of the body.The original scale formula is not recognizable.

dIscussIon
The general habitus of the studied fossils, together with the presence of dorsal limb of posttemporal firmly bound to the epioccipital (although this character is inferred based on superficial conditions and not directly observed), the ossification of the medial pelvic process distally, and the presence of true spines in the dorsal and anal fins, allows us to classify the fossils into the subsection Acanthomorpha Rosen, 1973(see Willey & Johnson 2010).This enormously diversified Teleostei crown group is in constant center of interest (e. g. Johnson & Patterson 1993;Willey et al. 2000;Davesne et al. 2016).Even after extensive research activities, a number of issues are not fully solved until today.
We are aware of the problems associated with the classification of fossil percoidei fishes, and these issues have been summarised with main points and references by Bannikov & Carnevale (2005).While the fossils at hand lack any direct characters that can be unequivocally used for classification into any of the acanthomorph orders, such as precise meristic characters or the distribution of predorsals or caudal skeleton, we applied a procedure based on the utilization of the maximal number of available characters.This procedure involves considering not only phylogenetically significant characters but also maximizing the use of available morphological features for a more comprehensive understanding of the fossils (Murray & Stewart 1999).

Familial and subfamilial classification of the fossils
The morphology of the opercle with two spines allows for the exclusion of certain groups characterised by a different number of opercular spines.For example, Serranidae possesses three opercular spines (Gosline 1966;Johnson 1983) and Latidae has a single opercular spine (Otero 2004).The opercle with two spines observed in the fossil specimens shares at least partial morphological similarity with opercles of some taxa from various families, including Percilidae (Percilia:  The family Percilidae, represented by two species within the single genus Percilia, is restricted to freshwater environments in South America and was formerly included in the family Percichthyidae (Nelson et al. 2016).Although the morphology of the opercle in Percilidae is superficially similar to that of our fossils, the upper spine of percilid opercle is not supported by thickened skeletal tissue, the ventral margin of the preopercular horizontal limb is smooth (with small spines developed in its corner), the cranial roof is smooth, and the con- struction of the neurocranium in dorsal view is significantly different (for details see Arratia 1982: fig.95).Therefore, based on these distinctive features, Percilidae can be confidently excluded.
The family Howellidae comprises marine fish distributed in the Atlantic, Indian, and Pacific oceans, with 9 species in three genera (Nelson et al. 2016).Prokofiev's detail osteological studies (2007a, b) have revealed different morphological characters in opercle and preopercle bones compared to our fossils.For example, in Bathysphyraenops the shape of the dorsal lamina and the absence of concave notch in between dorsal and ventral spine and in Howella, the shape of dorsal lamina, multiple spinulation in the distal part of the ventral opercular spine, and the significantly smaller concave notch in between dorsal and ventral spines.Additionally, the spinulation is in different rate and restricted to the corner of the bone in both Bathysphyraenops and Howella.Furthermore, the neurocranium in dorsal view shows differences in construction (Prokofiev 2007a: fig. 1;Prokofiev 2007b: fig.1).All these characters prevent unifying the fossils with the family Howellidae.
The family Acropomatidae encompasses marine fish distributed in Atlantic, Indian, and Pacific oceans, classified with 31 species in seven genera (Nelson et al. 2016).However, Synagrops, Parascombrops, Caraibops, and Kaperangus have been more recently classified within a separate family Synagropidae (Fricke et al. 2020;Smith et al. 2022).The opercles of Acropomatidae possess a more or less thickened lower opercular spine, but the upper spine and the anterior margin of the bone are not reinforced by bonny tissue, or are reinforced to a much lesser degree than observed in the studied fossils (Katayama 1952: figs 1 and2, 1959: figs 6, 9; personal observation).Further, the preopercle spinulation is developed in a different size and distribution pattern than those in the studied fossils.The most significant difference is in the morphology of the dorsal surface of the neurocrania-the shape of some bones and their configuration is not fully comparable, dorsal surfaces of the cranium lack any sculptation, and neural canals are expanded and largely opened (Katayama 1952: figs 1 and2, 1959: fig. 15; personal observation).
The skeletal fossil record of this family is limited, with Acropoma lepidotum (Agassiz, 1836) described from the Eocene deposits of Bolca, Italy (Sorbini 1975;Bannikov 2014;Carnevale et al. 2014) 1908;Arambourg 1967).There is also a possible Acropomatidae gen et sp.indet (originally described as Epigonus (?), see Bannikov 2000).Acropoma lepidotum differs from the studied fossils for reasons described for Recent representatives (see above).The genus Cottopsis can be excluded due to its opercle, featuring one dominant spine, the absence of concave notch, fine serration around the corner of the preopercle, and distinct details of the frontal morphology (Bannikov 2002; TP, personal observation of the Iranian specimens in the MNHN.F).
The family Polyprionidae is a group of marine fish recorded in the Atlantic, Indian, and Pacific oceans.Historically, the genus Stereolepis was also included in this family (e.g.Nelson et al. 2016).We herein follow Smith et al. (2022), which separate the genus Stereolepis into a distinct family (see below).The opercular apparatus of the genus Polyprion was figured by Arratia (1982: fig. 48), which shows certain degree of similarity with the studied fossils.However, key differences are evident, as only the lower opercular spine in Polyprion is supported by a thorn, and the preopercular denticulation is clearly different.Furthermore, Polyprion has a smooth cranial roof (Arratia 1982).
The family Stereolepididae includes two species within a single genus, which was previously classified within Polyprionidae but is now recognised as a separate group with a different phylogenetic position (Smith et al. 2022).The opercle of Stereolepididae, characterised by two spines supported by thorns, a deep concave notch, and an enforced anterior margin, is fully comparable with the studied fossils (see Katayama 1959: fig.6C, Fig. 11B).The general shape of the preopercle and its denticulation also align with the studied fossils.The dorsal side of the neurocranium is relatively flat, with a low and short supraoccipital crest.The anterior margin of the frontals creates a parabolic shape, and the sphenotics are relatively well developed and ventro-laterally bifurcated.The relative articulation of the individual bones is fully comparable with the studied fossils (compare Figures 9 and 11B).Although dermal sculptation is not explicitly mentioned by any authors, at least partial presence of the same type of dermal ornamentation was recognised in some adult specimens (see Ho et al. 2021: fig. 2; personal observation in CAS-ICH 66626).The fossil record of the family is restricted to a single remain of an angular bone determined as Stereolepis sp., collected in the Pleistocene deposits of Atsumi group in Japan (Ohe & Moriwaki 1996).Due to lack of respective remains, a direct comparison with the studied fossil is not possible.
The family Lateolabracidae, represented by the single genus Lateolabrax, is restricted to inshore waters of Western Pacific (Nelson et al. 2016).It exhibits a generally similar morphology of the opercle with our fossils (although the complete bone is surprisingly thick), but the horizontal arm of the preopercle in Lateolabrax is ventrally ornamented by several large spines (rather than average denticulation observed in the studied fossils).Additionally, the neurocranium in Lateolabracidae is smooth, generally elongated and differs from the studied fossils in other details (see Katayama 1959;personal observation).
The family Sinipercidae is a group of mainly freshwater fish found in Eastern Asia, with two Recent genera and 12 species (Fang & Chong 1932;Nelson et al. 2016).Several fossil taxa (one extinct genus) of this family have been described from the Lower Miocene freshwater deposits of Korean Peninsula (Nam et al. 2023) andChina (Chen et al. 1999), Middle Miocene of Japan (Yabumoto 2020;Yabumoto & Uyeno 2009), and Middle Miocene marine deposits of Japan (Yabumoto & Uyeno 2000).The fossils under consideration share relatively high similarity with sinipercids, but Siniperca can be excluded due to the lack of support for the opercular spines, large spines at the ventral margin of the horizontal preopercular ramus (in some respects reminding preopercle of Lateolabrax -compare figs 6E and 6D in Katayama 1959), and a generally elongated neurocranium (Katayama 1959: fig.17C).Similar arguments are valid for the exclusion of the extinct genus Inabaperca (Yabumoto & Uyeno 2000).The genus Coreoperca shows a generally very similar opercle, but it lacks support for opercular spines, with both spines being more or less the same size (Katayama 1959: fig. 6F;Chen et al. 1999: fig. 5;Yabumoto & Uyeno 2009: figs 2, 3;Nam et al. 2023: fig. 5), and a short neurocranium with a rather narrow anterior portion of the frontal bones (more concave in the studied fossils; Katayama 1959: fig.17C).Finally, the dorsal surfaces of the neurocrania are obviously smooth.
The family Symphysanodontidae, with two genera, is distributed in tropical and subtropical parts of the Pacific, Indian and Atlantic oceans (Kimura et al. 2017).The opercle with two spines (e.g.Johnson 1981;Kimura et al. 2017) is generally similar, but not supported in the same way as in the studied fossils (the lower spine is supported by an axis thorn, but not the upper one, and the distal part of both spines is thickened by stronger ossification).The preopercle is serrated in a different way, and the dorsicranium lacks typical sculptation (personal observation).
The family Moronidae, with six Recent species within two genera, inhabits brackish, freshwater, and marine coastal areas of North America and Europe (Nelson et al. 2016).The opercle is highly similar, with both spines more or less supported by thorns, but the preopercle shows significant differences (in Moronidae, the horizontal ramus is somehow longer and some larger spines are distally bifurcated ;Micklich 1988: fig. 14a; personal observation), and the dorsal side of the neurocranium is proportionally different, and its surface is smooth (personal observation).Two opercular spines in the extinct Oligocene genus Beaumontoperca Gaudant, 2000 lack thorny supports, and preopercular spines in the ventral section of the bone are bifurcated similarly as in other moronids (Gaudant 2000: fig. 6).
The studied specimens were also compared with Eocene European percomorphs of uncertain systematic position possessing similar morphology of the opercle with two spines.These are: Anthracoperca Voigt, 1934, Rhenanoperca Gaudant & Micklich, 1990, Palaeoperca Micklich, 1978, and Pararhenanoperca Micklich & Wuttke, 1988 (the number of opercular spines is not entirely clear in the latter genus, but noted below too).The results are as follows: 1.Two opercular spines in the genus Anthracoperca are in obviously different condition than in studied fossils (Micklich & Gaudant 1989: fig. 9), and neither a smooth preopercle nor dorsicranium is associable with studied specimens.2. Similarly, the genus Rhenanoperca also showed different morphology (Gaudant & Micklich 1990).3. The genus Palaeoperca possesses a preopercle with a completely different morphology of posterior spinulation of the vertical preopercular ramus (Micklich 1978;2205).4. The number of opercular spines is not clear in the genus Pararhenanoperca Micklich & Wuttke, 1988, but the preopercle is smooth, same as the neurocranium surface (Micklich & Wuttke 1988).
The provided brief overview, based on the comparison of available parts of the opercular apparatus (opercle and preopercle) and the dorsal side of the neurocranium, suggests a similarity with the family Stereolepididae, the group recently established by Smith et al. (2022).This family includes two species combined in a single genus-Stereolepis gigas Ayres, 1859 and S. doederleini Lindberg & Krasyukova, 1969-both inhabiting the subtropical part of the Pacific Ocean (Froese & Pauly 2023).Boulenger (1895) initially included Stereolepis in his Serranidae, and his diagnostic section for the genus included various characteristics that align with the studied fossil.For instance, the very small scales, a serrated preopercle with few strong antrorse spines on the lower border (which disappears in larger individuals), an opercle with two spines, a spinous portion of the dorsal fin much longer than the soft one (with XI+10), a shortbased anal fin with III+8, and a pectoral fin with 18 to 19 rays.Boulenger (1895) also mentioned strong anal spines, with the first shortest, second and third equal and about as long as the second dorsal fin spine for S. gigas.
Although there are some differences, such as in meristic counts of some fin rays, these are interpreted as resulting from inadequate preservation rather than reflecting the natural anatomical state.Further details that could confirm the classification, such as the precise predorsal formula, dorsal fin formula (see also Katayama 1960a, b), and other features, are challenging to determine due to insufficient preservation.The number of vertebrae (12+14), together with 9 to 12 dorsal fin spines and the presence of supramaxilla, would be crucial for confirming our classification within the family Stereolepididae (see family diagnosis in Smith et al. 2022).
Due to the absence of most parts of the viscerocranium and an articulated postcranial skeleton, the comparison of Stereolepis arcanum with its congeners is limited to the neurocranium.Stereolepis arcanum shows a shorter ethmoidal region, representing maximally 50% of the orbital length (in contrast to S. gigas and S. doederleini, where the ethmoidal and orbital areas are practically equal; personal observation) and much better-developed sphenotics, with the depth of this bone in lateral view approximately half of the ethmoidal depth (this bone is significantly smaller in both other species; personal observation).Particularly noteworthy is the strongly developed "pit and ridges" ornamentation on the dorsal surface of the frontal bones in S. arcanum, while this structure is only poorly developed in both other species (Ho et al. 2021: fig. 2B; personal observation).

Palaeobiological notes on the studied specimens
The stereolepids, particularly S. gigas, are known for their large size and are often sought-after catches by fishermen.Although their size might suggest that they can live up to a hundred years, recent studies based on S. gigas have revealed slightly lower ages, with the maximal recorded age being about 76 years (Allen & Andrews 2012;Hawk & Allen 2014).Data for S. doederleini are practically missing (noted in the latter publication).Given the size of the fossil specimens, it is reasonable to infer that they do not belong to fully grown and most probably non-adult individuals.Consequently, it seems reasonable to expect that some morphological characters and proportions may be different from those recorded for large, fully grown individuals.Such on-togenetic condition is recorded in various groups (see e.g.Cloutier 2010;Přikryl 2017).
Any morphological data for non-adults of Stereolepis are anecdotal and restricted to larval and early post-larval developmental stages (Shane et al. 1996;Ladin 2022 and references therein).However, Coufer (2017) and Benseman & Allen (2018) presented three phases of young-of-the-year Stereolepis gigas, which include black (10-21 mm TL), brown (23-33 mm TL) and orange (41-186 mm TL) phases.Based on the estimated SL of the studied specimens, it seems reasonable to compare them with individuals in the orange phase.However, the figures provided by Benseman & Allen (2018: figs 1, 2C) do not allow for the verification of details, and they certainly do not enable the evaluation of skeletal characters.Nevertheless, the available information suggests comparable proportions of spinous and soft parts of the dorsal fin, as well as proportions of individual spines in the anal fin.Such an assumption would lead us to the interpretation of the fossil specimens as individuals younger than one year.This interpretation is also supported by the absence of the growth rings at the vertebral centrae and opercle bone.Noichi et al. (1990) and Benseman et al. (2019) documented the behavior of juvenile Stereolepis, particularly their distribution in shallow and soft-bottomed habitats immediately below the wave base.Benseman & Allen (2018) mentioned that individuals in the orange phase were distributed "almost exclusively" at depths of 6 to 10 meters.These observations are valuable for contextualizing the environment in which the studied fossils might have lived.However, it is essential to consider the sedimentology and the accompanying fossils.The relatively high degree of disarticulation, incompleteness, and partial abrasion observed in the fossils may indicate a highly dynamic environment rather than re-deposition.
The genus Stereolepis, which is currently restricted to the Eastern and Northwestern Pacific (Froese & Pauly 2023), has a notable southern extension with the recent discovery of Stereolepis doederleini in Taiwan by Ho et al. (2021).It is necessary to mention that such southern occurrence is considered relatively exceptional in the present-day context.Although data on climate change in this region during the Pliocene are scarce, the past occurrence of Stereolepis in southern regions is foreseeable, es-pecially considering the prevailing oceanic currents and the topography of the ocean floor around Taiwan (Jan et al. 2002).
The contrast between the fossil and modern distribution of Stereolepis in Taiwan during the Early Pliocene adds an intriguing insight into our understanding of the historical biogeography of this genus.These palaeobiological findings highlight the importance of continued palaeontological exploration in the underrepresented tropical-subtropical West Pacific.

conclusIons
The semi-articulated specimens from the Lower Pliocene deposits of the Upper Kueichulin Formation in northern Taiwan are identified as a new species of the family Stereolepididae, Stereolepis arcanum n. sp., based on the morphology of the neurocranium.While some characters are partially incomplete, they are consistent with the features observed in the family Stereolepididae.However, they also distinctly differentiate the specimens from previously described species.The estimated age of the specimens is less than one year, based on comparisons with the size of Stereolepis gigas.The discovery of the specimens represents the second record of this group globally and highlights the importance of ongoing fossil exploration to enhance our understanding of fish palaeobiodiversity and evolutionary history in the Indo-West Pacific region, particularly during the Neogene.
Taiwan and the Czech Republic -the once thrived ichthyofaunas to T. Přikryl and C.-H. Lin, and the National Science and Technology Council, Taiwan (Grant No. 110-2116-M-001-009, 111-2116-M-001-033, 112-2116-M-001-017-MY3) and Academia Sinica, Taipei, Taiwan to C.-H. Lin.TP is supported by the Institute of Geology of the Czech Academy of Sciences (RVO67985831).We thank Alison Murray (University of Alberta) and Werner Schwarzhans (Natural History Museum of Denmark) for their comments and review of the earlier version of the manuscript.

Fig. 1 -
Fig. 1 -The geological setting of the acropomatid fossils.A) Location of the study area; B) geological map of the study area; C) geological column and stratigraphy of the study area (modified from Pan et al. 2015; Chen 2016); D) photograph of the locality; E) exposure where the fossil nodules were found.Red stars represent the sampling site.
Institutional abbreviations: ASIZF -Biodiversity Research Museum, Academia Sinica, Taiwan; CAS-ICH -California academy of Sciences, Ichthyology Collection; GLU -Institute of Geology of the Czech Academy of Sciences in Prague, the Czech Republic; MNHN.F -fossil fish collection, Muséum National d'Histoire Natu-relle, Paris, France; NMMB -National Museum of Marine Biology and Aquarium, Pingtung, Taiwan; NMMST -National Museum of Marine Science and Technology, Keelung, Taiwan.

Fig. 5 -
Fig. 5 -Detail CT examination of holotype specimen of Stereolepis arcanum n. sp.NMMST 20190-3001.A) Right external surface of the skull showing spinulation of the preopercle and preserved supraneural (marked by arrows); B) right inner scan showing the spinulation of the lower part of the preopercle bone (individual spines marked by arrows); C) right inner scan showing the general shape of the neurocranium in lateral view; D) inner scan showing the general shape of the neurocranium in dorsal view.

Fig. 6 -
Fig. 6 -Neurocrania of Stereolepis arcanum n. sp.All specimens are in dorsal view.A) Holotype specimen NMMST 20190-3001; B, C) paratype specimen NMMST 20190-3005.Note the third neurocranium preserved in NMMST 20190-3005 and the isolated skull in ASIZF0100722 are too incomplete to provide any visual information and are not figured here.