Intraspecific Variability in Paradoxidid Trilobites from the Purujosa Trilobite Assemblage (Middle Cambrian, Northeast Spain)

Eccaparadoxides? pradoanus and Eccaparadoxides mediterraneus are both widespread trilobite species described from the middle Cambrian of the Mediterranean region. Analysis based on a pooled sample of 500 specimens demonstrates that many of the characters that have been used to define these species show continuous variations, some of which are related to ontogeny. In addition, morphometric analyses of metric characters show that the two species cannot be distinguished on the basis of these characters either. Many of the characters studied herein are widely used in definitions and descriptions of other paradoxidid species, which suggests that the taxonomic classification of other paradoxidids may be oversplit.


Introduction
Although "paradoxidid" trilobites are among the most familiar middle Cambrian fossils, many details of their evolution and development remain poorly known.This is largely because most species within the group have been erected on the basis of a few, often isolated sclerites as it is also the case with many other Cambrian trilobite species (Hughes 1994).Moreover, trilobites of this age quite commonly displayed high degrees of morphological plasticity (Rasetti 1948;Mc-Namara 1986;Webster 2007), which further hinders cogent species recognition.Hughes and Labandeira (1995) demonstrated that taxonomic studies based on small numbers of specimens can lead to excessive taxonomic splitting.Thus, the combination of high plasticity and low numbers of specimens used in the erection of species may have contributed to the recognition of many of the 144 paradoxidid species and subspecies erected over the last one hundred and fifty years (Geyer and Landing 2001).For example, those known only from a few isolated sclerites include Paradoxides rugulosus Bergeron, 1889, Paradoxides mediterraneus Pompeckj, 1901; Paradoxides asturianus Sdzuy, 1968; Pararadoxides? enormis Sdzuy, 1968, Eccaparadoxides granulosus Courtessole, 1973, Paradoxides (Eccaparadoxides) sulcatus Liñán and Gozalo, 1986, Eccaparadoxides kozaki Kordule, 1999.In addition some species have been re-described on the basis of an insufficient number of specimens to address the potential problem of intraspecific variation (e.g., Eccaparadoxides acadicus [Matthew, 1883], Eccaparadoxides lamellatus [Hartt in Dawson, 1868], [Kim et al. 2002]).
Good stratigraphic and taphonomic control is essential for making an informed assessment of taxonomy and intraspecific variation.In addition to being based on small numbers of specimens, many Paradoxides species also lack adequate stratigraphic and taphonomic control.Furthermore, ontogenetic studies are necessary in order to assess whether observed morphological differences might relate to differences among growth stages.
Ten genera were considered valid within the family Paradoxidae Hawle and Corda, 1847 (see Dean and Rushton 1997).This article focuses on the morphology and variation of two species of Eccaparadoxides Šnajdr, 1957, a genus found in the Mediterranean area, and also in Bohemia, Siberia, and Avalonia.These species are Eccaparadoxides mediterraneus (Pompeckj, 1901) and Eccaparadoxides?pra doanus (Veneuil and Barrande in Prado et al., 1860).The primary features used to differentiate these species are (i) the presence of two glabellar furrows in E.? pradoanus and four in E. mediterraneus; (ii) a pygidium without posterolateral spines in E.? pradoanus but with more or less well developed posterolateral spines in E. mediterraneus, and (iii) a homonomous condition in E.? pradoanus but a homonomous or heteronomous conditions shown in E. mediterraneus.
These species were studied because they are abundant in the Purujosa trilobite assemblage (Northeast Spain) and both have a broad geographic range (Spain, France, and Sardinia) and good stratigraphical control in various sections in the Iberian Chains as well in the Cantabrian Mountains (North Spain).Nevertheless, to date existing collections, made from a single level or one locality (and including both disarticulated and articulated material), are insufficient to permit detailed analysis of intracollectional variation.However, the present study has been able to address this deficiency.Furthermore, most of the collections from other localities show a size sorting with either large or small specimens present, what prevents recognition of ontogenetic variability throughout the trilobites' growth history.The Purujosa trilobite assemblage (Northeast Spain) offers a large number of specimens, many of which are articulated and enrolled.Trilobites are typically preserved as internal and external moulds and often preserved through mineral replacement as chlorite, and with a relatively low degree of tectonic deformation (Esteve et al. 2010(Esteve et al. , 2011(Esteve et al. , 2012;;Zamora and Esteve 2010).Moreover, this assemblage offers a large size range Fig. 1. A. Geological setting of the Purujosa trilobite assemblage in the Iberian Chains (modified from Gozalo and Liñán 1988).B. Geological setting, showing pre-Hercynian outcrops and the Iberian Chains in NE Spain.C. Composite column with middle Cambrian formations and Mediterranean substages showing the stratigraphical distributions of Eccaparadoxides mediterraneus (Pompeckj, 1901) and Eccaparadoxides pradoanus (Verneuil and Barrande in Prado et al., 1860).Tectono-stratigraphical zones of Iberian Peninsula: CZ, Cantabrian Zone; ELAZ, East Lusitanian-Alcudian Zone; GCZ, Galician-Castilian Zone; OMZ, Ossa-Morena Zone; SPZ, South Portugal Zone; WALZ, West Asturian-Leonese Zone.
extending from late meraspids to large holaspids of both E. mediterraneus and E.? pradoanus.This material allowed a comprehensive analysis of variation in discrete, ordinal, and metric characters.The type specimens of both species were included in the analysis.

Stratigraphic and geological setting
Both Eccaparadoxides mediterraneus and E.? pradoanus are commonly found as incomplete and deformed fossils (see Liñán and Gozalo 1986;Dies Álvarez et al. 2010) throughout the late Caesaraugustian to Languedocian (ca.506-507 Ma) (Fig. 1A) in the Murero Formation (Liñán et al. 1993) and the Borobia Formation (Álvaro 1995) in the Zaragoza and Soria provinces of Spain.The rich assemblage of articulated material occurs in the top of the Murero Formation at the Purujosa locality, in northeast Spain (Fig. 1B).It also bears a large number of articulated sclerites of other trilobite species (Solenopleuropsis thorali Sdzuy, 1958, Solenopleuropsis marginata Sdzuy, 1958, Conocoryphe heberti Munier-Chalmas and Bergeron in Bergeron, 1889, Schopfaspis?graciai nomen nodum Esteve et al. 2012), showing varied degrees of enrollment and also co-occurs with an unusually high diversity of echinoderms (Zamora 2010;Zamora andSmith 2010, 2012;Esteve et al. 2010Esteve et al. , 2011Esteve et al. , 2012)).The Purujosa trilobite assemblage occurs within a one meter thick red mudstone with a low degree of bioturbation (ichnofabric index 1-2 of the Droser and Bottjer 1986 scheme).The taphonomic and lithological data suggest that this benthonic community was entombed by sudden burial events that allowed preservation of a high number of articulated specimens (Esteve et al. 2011).

Material and methods
More than 500 specimens of Eccaparadoxides were studied using a Leica M165C stereomicroscope equipped with an eye piece scale (each division equal to 0.034 mm).Specimens larger than a cranidial length of 10.0 mm were measured using calipers.ImageJ software (Abràmoff et al. 2004) was used to analyze images of the specimens and to derive both linear and angular measurements from these; this software allows measurements of gradational features such as palpebral furrows or border furrows and minimized the errors involved of these linear variables.
Firstly, analysis of those characters previously used to describe these species was undertaken.Some observed morphological variations, such as the presence or absence of features (e.g., a particular pair of glabellar furrows) belong to a class called nominal characters (Hughes 1994: 6).Other characters (e.g., the specific number of thoracic segments, number of terraces lines, etc.) show discrete variation along a ranked scale, and are termed ordinal characters (Hughes 1994).In addition, bivariate relationships between linear variables were calculated using the reduced major axis (RMA) approach using the Log10 values of the original linear dimensions measured in millimeters.The analyses were carried out using the RMA 1.17 software written by Andrew Bohanak of San Diego State University, this software provides error estimate which is calculated using three methods: resampling was completed 1000 times in each case to permit the calculation of confidence intervals for both the slope and each RMA.When preservation permitted, each of the 14 linear dimensions assessed was measured on 167 cranidia and 113 pygidia: 8 cranidial (3 length measurements obtained parallel to the sagittal axis and 5 width measures obtained orthogonal to the sagittal axis) and 6 pygidial (3 length measures parallel to [articulating half-ring is omitted] and 3 width measures orthogonal to the sagittal axis) (Fig. 2).For a synoptic view of some results obtained with RMA a principal components analysis (PCA) was performed using the PAST software package (Hammer et al. 2001).In order to assess deformation PCA helps to examine whether the effects of deformation correlate with a principal component axis.PCA was based on the correlation matrix for 4 linear measurements in 117 cranidia, and 5 linear measurements in 46 cranidia, and on 6 linear measures in 98 pygidia, using only those specimens that preserved all these characters.These analyses were traditional in that they used data from linear measurements, rather than a landmark based approach.This approach was deemed appropriate due to the compacted nature of the material at hand, with more sophisticated morphometric approaches exceeding the quality of the material available.Although it would be possible to apply geometrics morphometrics to some anatomical parts such as the glabella, interpreting the results would likely be very difficult and not significant (see Hughes 1999).
All the specimens from the Purujosa trilobite assemblage are housed in MPZ.The lectotypes of Paradoxides pradoanus are housed in EM.

Nominal and ordinal characters of paradoxidid trilobites
Glabellar furrows.-Glabellarfurrows have been an important feature for description of paradoxidids.For example, Šnajdr (1957) erected the genera Acadoparadoxides and Eccaparadoxides based, among other features, their diagnoses on the number of glabellar furrows as a distinguishing feature: two pairs in Acadoparadoxides and four pairs in Eccaparadoxides (exclusive of sulcus occipitalis [SO] in the original description).Within the genus Eccaparadoxides, the number of transglabellar and discontinuous furrows has been an important feature in the determination of Eccaparadoxides species (Šnajdr 1957, 1986;Sdzuy 1961Sdzuy , 1968;;Courtessole 1967Courtessole , 1973;;Liñán and Gozalo 1986;Courtessole et al. 1988;Kordule 1999;Dies Álvarez et al. 2010).
Eccaparadoxides mediterraneus bears two transglabellar furrows and two pairs of discontinuous furrows whereas E.? pradoanus bears only two transglabellar furrows (sensu Sdzuy 1961).All specimens of Eccaparadoxides from the Purujosa trilobite assemblage have at least one pair of glabellar furrows (S1) while some specimens have all the other furrows up to S4, additional pairs (S2, S3, S4) may be absent in poorly preserved specimens due either to taphonomic loss during flattening, or due to a natural variation in the expression of weakly incised anterior furrows (see Hughes 1994) (Figs. 3, 4).Accordingly, within the Purujosa assemblage there are different patterns of glabellar furrows (Fig. 3).Sdzuy (1961) described the S1 to be curved backwards medially in E. mediterraneus as well as E.? pradoanus and Dies Álvarez et al. ( 2010) also described S1 as curved backwards in E. mediterraneus.None of these authors took into consideration a possible angular variation of S1 and SO.The specimens from the Purujosa assemblage show some specimens with convergent SO (strongly curved forwards) and S1 (strongly curved backwards medially) whereas in These patterns are seen in the 95% of the specimens although it is likely that other patterns can be found.Note that the ornamentation is strong in A and weaker in D and absent in E and F.
others there is no evident convergence of either structure (Figs.3-5).Therefore, the angular relationships of SO and S1 to the sagittal axis were measured in order to document their range of variation (Fig. 6).Angular measurements on S1 show high variability in both species: the angle in SO ranges from ca. 0º and 39º and in S1 it ranges from 10º to 35º.This variability occurs in both species.The correlation coefficient between the furrow angle and glabellar length is very low (r = -0.13,ρ = 0.48 for S1 and r = 0.14 ρ = 0.56 for SO) suggesting independence between the overall size and both angles.The analysis suggests continuous variation in this feature.Therefore there is a continuity of character states between specimens with straight SO and S1 (Figs. 4F, G, 5A, I) and specimens with strongly convergent SO and S1 that almost meet at the medial plane (Fig. 5C, E).Although deformation is very weak the angular relationships between pair of furrows have been measured in those specimens without evidences of deformation, in order to avoid any influence of deformation.The number of furrows has been used in the diagnosis and descriptions of some Eccaparadoxides species (Sdzuy 1961(Sdzuy , 1967;;Courtessole 1967Courtessole , 1973;;Liñán and Gozalo 1986;Courtessole et al. 1988;Kordule 1999;Kim et al. 2002;Dies Álvarez et al. 2010) and Gozalo (1986) described both species from the Murero locality.In all of them, the specimens show the same pattern of variability.Therefore, it is likely that this variability is intraspecific among different populations but also locally enhanced by differences in taphonomic history.
Those specimens with only two pairs of glabellar furrows have been assigned to E.? pradoanus by Liñán and Gozalo (1986).In spite of the holaspid specimens figured that article have two pairs of glabellar furrows (S1 and S2), the authors assigned the species to the genus because Sdzuy (1961: 322, pl. 18) illustrated a few associated meraspid specimens bearing at least three pairs of glabellar furrows.
They argued that during the ontogenetic sequence this trilobite lost one pair of glabellar furrows.Hughes (1994: 11) suggested that the number of glabellar furrows is variable within the holaspid Dikelocephalus minnesotensis Owen, 1852 and the number of glabellar furrows seems to be partly related to size.However, although specimens within Sdzuy's (1961) collection show a decreasing number of glabellar furrows with increasing size, no size-related trend was detected within the Purujosa trilobite assemblage.In an attempt to resolve this paradox, more specimens were collected from Sdzuy's (1961) locality (Los Barrios de Luna, Cantabrian Mountains, N. Spain).These show the same pattern of variability in the glabellar furrows as that of the Purujosa assemblage.This suggests that the ontogenetic pattern apparently seen in Sdzuy's (1961) material may have been an artifact of small sample size.On other hand, Verneuil and Barrande in Prado et al. (1860) described "Paradoxides pradoanus" from the locality of Sabero in the Cantabrian Mountains, close to Los Barrios de Luna (Fig. 7).These specimens preserved in mudstone have a glabella with two pairs of transglabellar furrows and two pairs that are discontinuous, what corresponds to Šnajdr's (1957) definition of Eccaparadoxides.Thus, I suggest that assignment to Eccaparadoxides is correct.However, the presence of two pairs of glabellar furrows as a single character does not prove that the specimens belong to Acadoparadoxides because other additional features demark the difference between these genera (see Geyer and Landing 2001).Therefore both species appear to show the same variability in the number of glabellar furrows.
Ornamentation.-The specimens of Eccaparadoxides within the Purujosa assemblage show three types of ornamentation: granules, terraces lines, and genal caeca.Ornamentation is a feature sometimes used in the diagnoses or descriptions of species of Eccaparadoxides.Sdzuy (1968: 93) used the ornamentation to distinguish between E. asturianus and E. lamellatus.Courtessole (1973: 131) introduced E. granulosus (Courtessole, 1973) on the basis of specimens bearing granules.Šnajdr (1986: 171) erected E. rohanovicus Šnajdr, 1986 using specimens with the test covered by fine granules intercalated with the terraces lines on the anterior border.Kim et al. (2002: 841) also diagnosed E. lamellatus based coarsely granulose sculpture on the glabella.Likewise, Dies Àlvarez et al. (2010: 103) distinguished E. pusillus (Barrande, 1846) and E. mediterraneus on, among other features, the lack of granules on the exoskeleton of E. mediterraneus.
The preservation of specimens from the Purujosa assemblage does not permit a precise analysis of the variation in granulation.Nevertheless, a transition between specimens with granules and those without granules is recognizable (Figs. 3, 8) and ornamentation is more conspicuous in the external mold when it is preserved.In addition some specimens of E.? pradoanus as well as of E. mediterraneus show, if preserved, the same kind of ornamentation.Because all specimens come from the same bed, it appears likely that the presence or absence of granules in specimens found in mudstone may be related to preservation.Likewise, it is likely that different species show difference types of granules.Therefore this feature may be used to differentiate among Eccaparadoxides species which are preserved in limestone.By contrast when such trilobites are preserved in mudstone the ornamentation should be used only after careful consideration.
Terrace lines are very common in specimens from the Purujosa assemblage.As it is the case with granules, the presence or absence of terraces lines appears to be related to preservational processes.The terraces lines are located on the anterior border and on the rostral plate, on the lateral border of the cephalon, along the genal spines and on the doublure, on the dorsal surface and the doublure of the pleural tips, and on the dorsal surface and the doublure of the pygidium.All specimens from the Purujosa assemblage lack terrace lines on the glabella.I carried out an analysis of the relationship between size and number of terrace lines (Fig. 9).This analysis indicated a high correspondence between size and number of terraces lines (Spearman's D correlation ρ = 0.0005).In addition, throughout the growth the distance between terrace lines seems to diminish towards the border.
Finally, genal caeca are preserved in a few specimens of

E. pradoanus and E. mediterraneus (Figs. 4F, H, 10A
).The very weak genal caeca contrast with the markedly developed genal caeca of E. rouvillei (Miquel 1905).Therefore genal caeca cannot be used to differentiate Eccaparadoxides species because the presence or absence of this feature is both biologically and taphonomically controlled.
Macropleural and micropleural spines.-Eccaparadoxidespradoanus has a homonomous thorax with all segments showing a similar shape.However, E. mediterraneus shows two morphotypes.Some specimens bear a homonomous thorax, others a heteronomous thorax with three different batches of segments.Because the inner part of all segments has an identical shape in all thoraxes, the terms macropleural and micropleural segments used in previous descriptions are inadequate in this case.Palmer (1998) suggested a terminology for different olenellids thoracic segments.He suggested term "macrospinous" for segments in which the inner part does not change but which bear longer pleural spines.Here, I follow this terminology and additionally I use the term "microspinous" for specimens with smaller pleural spines.All heteronomous examples have two macrospinous segments in the first batch; however, in the second and the third batches the number of segments varies between 8 and 3 microspinous segments in the second batch and between 6 and 13 regular segments in the third batch.Macrospinous segments in Eccaparadoxides have been reported for different species.Šnajdr (1957) suggested that holaspid specimens of E. pusillus lack macrospinous segments but simultaneously illustrated meraspids that bore them.Meraspid specimens of E. pusillus bear two macrospinous segments and the rest of the pleural spines are small  (Boeck, 1827) or Hydrocephalus minor (Boeck, 1827) (see Šnajdr 1958: pls.12, 23).Liñán and Gozalo (1986: 54-59) also reported specimens of Eccaparadoxides (E.sequeirosi [Liñán and Gozalo, 1986] and E. mediterraneus), which varied in showing the homonomous and heteronomous conditions, and other such as E. pradoanus with only the homonomous condition.They suggested that E. pradoanus is not phylogenetically related to E. mediterraneus because of its homonomous condition.Nevertheless they also suggested that the macrospinous length decreased during ontogeny.Liñán and Gozalo (1986: 57) and Gozalo et al. (2003) suggested that species with heteronomous and homonomous conditions provide examples of intraspecific dimorphism, which they interpreted to be sexually related.
Eccaparadoxides from the Purujosa trilobite assemblage also provided specimens with either the homonomous (Fig. 5) or with the heteronomous condition (Fig. 10).Within the Purujosa trilobite assemblage small specimens (ca.15 mm long) bear macrospinous segments followed by 7-8 microspinous segments.However, not all of the smaller specimens bear macrospinous segments, although protaspid and majority of the early meraspid stages are absent in the sample.On the other hand, few medium-sized specimens (ca. 5 cm in thoracic length) are heteronomous, while most are homonomous.However, no heteronomous large sized specimens of Eccaparadoxides (>10 cm in thorax length) have been reported from Purujosa or elsewhere, although other big Bohemian Paradoxides and Hydrocephalus retain a slightly elongated macrospinous segments (Adrian Rusthon, personal communication 2011).The Purujosa assemblage sample does not include a high number of specimens with a heteronomous thorax (<5%).However, specimens from other localities show that the macrospinous segments reduced the length while the rearward microspinous segments increased in relative length through ontogeny (Fig. 11).In addition, some specimens bear 8 microspinous segments with two large macrospinous segments and when the macrospinous segments are reduced, the length of last microspinous segment is increased.Eventually, this micropleural spine disappears being transformed into a regular pleural spine, in the next molt so that the trilobite is reduced in the relative length of the two macrospinous segment and increases in the relative length of the microspinous segments; a procedure that may have been repeated until the last trace of the microspinous condition has disappeared.Therefore, after several molts a specimen with the heteronomous condition, Eccaparadoxides mediterraneus may have transformed into a form with a homonomous trunk, E. pradoanus (Fig. 11).Different delays on the modifications of the heteronomous condition during the growth of the   (Figs. 5,10).This variation has been compared with cranidial and pygidial length (Fig. 12) in order to assess whether variation is ontogenetically related.Spearman's rank correlation does not indicate a correlation between the number of segments and either pygidial length (n = 27, ρ = 0.29) or cranidial length (n = 23, ρ = 0.13), which does not suggest a clear relationship between size and the number of thoracic segments among holaspids specimens from the Purujosa assemblage.In addition the smallest observed cranidia are distinctly larger than the smallest holaspid in species for which the meraspid-holaspid transition is well known (e.g., Hughes 1994;Feist and Lerosey-Aubril 2005).Moreover, the addition of segments during the holaspid growth appears to be unlikely (Fusco et al. 2004).Therefore, variability in the number of segments is apparently a holaspid intraspecific variation unrelated to size.Specimens of both species from other localities of the Iberian Chains, such as Murero, and specimens from different localities of the Montagne Noire (France) also show the same pattern of intracollectional variation.However, in these cases of articulated meraspid specimens it will be necessary to assess whether these variations could be related to ontogeny.
Variation in the number of thoracic segments has been reported in other paradoxidids such as Paradoxides davidis Salter, 1863 which has between 18 and 21 thoracic segments (see Bergström and Levi-Setti 1978: fig. 6).
Posterolateral pygidial spines.-Twoclearly different posterolateral pygidial spine morphotypes are present in the Purujosa trilobite assemblage (Figs. 2,13).The first morphotype lacks the posterolateral pygidial spines (E.pradoanus) while the second has posterolateral pygidial spines (E.mediterraneus).Liñán and Gozalo (1986) noted that posterolateral pygidial spines increase length through ontogeny.I agree with that observation however, within the Purujosa assemblage, an unusual pattern is observed.While in other trilobite taxa pygidial spines often tend to decrease in length during growth (e.g., Park and Choi 2009), the spines in Eccaparadoxides lengthen relative to the overall size during holaspid ontogeny.In addition, this growth does not take place continuously during the phase of the holaspid growth.I have calculated an estimate of the spinosity-degree (ratio of the exsagittal pygidial length to the sagittal pygidial length) and plotted this value against the maximum pygidial width (Fig. 14).The plot shows a high variation among specimens below 15 mm in the maximum pygidial width and a correlation between spinosity-degree and size among large specimens (exsagittal pygidial length >10 mm).
These results suggest that spine morphology undergoes a marked and most unusual change during holaspid ontogeny.Therefore there is a marked morphological difference between specimens with well developed spines and those trilobites without spines.It is important to note that pygidia show a size control throughout the Murero Formation attributed to sorting, some beds yielding large pygidia and others small pygidia only.Also there is size-sorting within the Purujosa assemblage, although intensive sampling provided an almost complete size-spectrum.Nevertheless, there is an absence of specimens between 10 and 15 mm maximum pygidial width which does not allow discriminating as whether the growth transition from non-spinose to spinose individuals took place gradually or whether a sharp inflection took place.If there were two phases, it might indicate that the pygidia of some specimens (i.e., those with small spines) experienced a heterochronous dwarfism.However, although this fact has been observed in trilobites from other localities from the Iberian Chains, the Cantabrian Mountains and the Montagne Noire, no complete collections are available yet from other single beds allowing independent test of this hypothesis.Even so, the data support the notion that all pygidia belong to a single species that displays a change in its mode of growth, rather than two or more different species.
Discussion.-The analysis of 6 nominal and ordinal characters shows a considerable variability within the sample from the Purujosa trilobite assemblage.The analysis indicates that only a single morphospecies may be recognized.In addition, all trilobites were collected from the same bed.Some characters provide evidence for ontogenetic control.One example is that the number of terrace lines per unit area increased during ontogeny.Also, there is correspondence between the size and the number of thoracic segments although this variation is not likely to be ontogenetically controlled.The number of glabellar furrows has doubtful size correspondence in the Purujosa assemblage and was likely subject to taphonomic control.Macrospinous and microspinous segments show high variability.Both homonomous and heteronomous forms have been found both in small as well as in large holaspid specimens and, despite the fact that size control is not evident, preliminary observations suggest post-displacement paedomorphosis.Although this variability is observed only in the material from the Iberian Chains, other characters such as the pygidial spinosity are highly variable among small specimens whereas large specimens tend to show a size-related trend towards increased spinal length.However, the heterochronic dwarfism hypothesis remains open.All these characters have been noted in the description of one or more paradoxidid species (Sdzuy 1961;Courtessole 1973;Liñán and Gozalo 1986;Dies Álvarez et al. 2010).In summary, this study indicates high intraspecific variation in all characters analyzed.These results suggest that individuals of Eccaparadoxides pradoanus and E. mediterraneus both belong to a single biological species.

Bivariate analysis: reduced major analysis (RMA)
Although analyses of nominal and ordinal characters show evidence of wide morphological variation, a biometric analysis of metric characters provides a basis for assessing whether consistent differences in aspects of shape exist.Bivariate analysis has been used to examine the relationship between pairs of variables through holaspid growth of cranidia and pygidia (Fig. 2B).
Cranidium.-Themorphological variation of each cranidial character has been assessed with respect to a standard measure for cranidial size.The standard measure has been chosen as the sagittal glabellar length due to axial features being relatively constant in trilobites (Palmer 1957;Hughes 1994).
Anterior border length.-Thegrowth of the anterior border is negatively allometric with respect to the glabellar length.There is considerable variation in the length of the frontal area with respect to glabellar length within the sample.The growth of the anterior border is negatively allometric at the 95% confidence level (Fig. 15, Table 1A).This suggests that the anterior border allometry is strong.Smaller specimens have a large frontal area with a long preglabellar field which is reduced in size and fades in large specimens (Figs. 4,10).
Palpebral lobe length.-Thegrowth of the palpebral lobe with respect to glabellar length is isometric (Fig. 16A).However, there is variation in the palpebral lobe in relation to glabellar length within the sample.The bootstrap analysis shows values of slightly negative allometry at the 95% and 99% confidence level, because large sample sizes are needed to detect it (Table 1B).
Palpebral lobe width.-Thegrowth of the transverse palpebral lobe width with respect to the glabellar length is slightly negatively allometric.There is slight variation in the palpebral lobe width related to glabellar length within the sample (Fig. 16B, Table 1C).However, the growth is isometric at the 95% confidence level.
Posterior glabellar and maximum glabellar width.-Themaximum and posterior glabellar width both vary with respect to glabellar length indicating isometric growth (Table 1D, E).There is a slight variation in both variables within the sample.However, the confidence intervals at 95% and 99% are quite wide (Table 1D).Because the specimens display slight tectonic deformation, the widely dispersed values may reflect distortion due to compaction, notably in the glabella, which if flattened in most of the specimens.The posterior glabellar width has been assessed with reference to the posterior border breadth and shows isometric growth at the 95% confidence level (Table 1F).These results suggest that there is no significant variation in the posterior and anterior glabellar width within the sample.The bivariate plot shows more dispersion among large specimens (Fig. 16C).
Frontal area width and posterior border width.-Thegrowth of frontal area and the breadth of the posterior border with respect to the glabellar length are isometric (Figs.16D, 17, Table 1G, H).There is high variation in both variables within the sample.There is a strong correlation between frontal area and glabellar length (r = 0.92, n = 63) and the RMA show isometric growth in the sample at the 95% confidence level.The correlation between the breadth of the posterior border and the glabellar length is slightly low (r = 0.86, n = 54) but RMA also show isometric growth in the sample at the 95% confidence level.
Anterior pygidial width.-Thegrowth of the pygidial anterior section is isometric at the 95% confidence level.However, there is a slight variation in the anterior pygidial width compared to the sagittal pygidial spine length within the sample.(Fig. 18A, Table 2A).The bootstrap analysis shows a slightly of negative allometry at the 95% and 99% confidence levels (Table 2A).This result suggests that the anterior pygidial width could be allometric.

Medium pygidial width (maximum)
.-The growth of the pygidial width at the midpoint of its length is isometric.There is variation in the medium pygidial width related to the spine pygidial length (exsagittal) within the sample.There is a strong correlation between the medium pygidial width and the sagittal pygidial length (r = 0.97, n = 107), and the RMA shows isometric growth in the sample at 95% confidence level (Fig. 18B, Table 2B).
Posterior pygidial width.-Thegrowth of posterior pygidial width is almost isometric.There is a correlation between medium pygidial width and the sagittal pygidial length (r = 0.89, n = 108) and the RMA shows isometric growth in the sample at the 95% confidence level (Fig. 18C).2D).There is a slight variation in the pygidial length related to the sagittal pygidial length within the sample.A strong correlation exists between the pygidial spine length and the exsagittal pygidial spine length (r = 0.99, n = 106) and the RMA shows almost isometric growth in the sample at 95% confidence level.

Pygidial length (sagittal).-The growth of the pygidial length is almost isometric (Table
Pygidial axial length.-Thegrowth of the pygidial axial length is isometric.There is a slight variation in the pygidial axial length related to the pygidial length within the sample. A strong correlation exists between the pygidial axial length and the sagittal pygidial length (r = 0.93, n = 108) and the RMA shows an almost isometric growth in the sample at the 95% confidence level (Fig. 18D, Table 2E).
Discussion.-The results show that variations of few characters are independent of size.Growth patterns indicate that most of the characters have isometric growth although others develop more or less clearly allometrically.Only the cranidial frontal area length shows strong allometric growth and a low degree of allometric growth took place for the palpebral lobe length.Nevertheless correlation among some bivariate characters such as the frontal area and the posterior border width show relatively low values suggesting highly independent-size variability in these characters.For the pygidium, only the anterior pygidial width shows a clear allometric development in the holaspid phase.This allometric growth is responsible for changes in the general morphology of the pygidium because of the allometric growth pattern followed by an overall morphological variation.Others variables such as the posterior pygidial width and the pygidium length show low RMA-slope values (Table 2C, D), suggesting a slight size-dependent variability and a continuous change in morphology.That recognition of both allometric and isometric growth and of size independent changes are important because show the morphological variation within the sample that suggest a single morphospecies.However, some of this variation could be related to compaction-related deformation.This varia-tion is detected in analyses performed with perpendicular measurements, which obtained low correlation coefficients (R <0.70, see Tables 1 and 2).On the other hand, Hughes (1999) suggested that large specimens suffered more deformations than smaller ones.There is a size threshold above which marked asymmetry is common in those trilobites preserved in mudstone.Within the sample the threshold occurs at glabellar sizes of >15 mm.

Multivariate analysis: principal component analysis (PCA)
Although the bivariate analysis suggests that both "traditional species" in fact constitute only one morphospecies, it is possible that (if viewed synoptically) multiple measurements could reveal the presence of separate clusters.Thus a principal component analysis (PCA) was employed to test this possibility.Material and methods are discussed above.Cranidia.-BecausePCA require complete data for each specimen considered in the analysis, two different PCAs were computed in order to assess results, both based on correlation matrices (Table 3A, B).The first analysis included all characters considered in the bivariate study except for the anterior glabellar width and the posterior border width (n = 46).The second analysis also excluded the frontal area width (n = 117).
The eigen-values (Table 4A, B) are an indication of how much of the variation is accounted for by each principal component, and the contributions of each variable to each principal component have also been calculated as the eigen-weights (Table 5A, B).The first and second components account for most of the variation within the sample in both analysis (94% and 96.8% respectively).All variables have positive eigen-values in both PCAs, and eigen-weights are roughly similar among all variables (Table 4A, B).This fact suggests that the first component reflects the overall size of the specimen, and this accounts for most of the variation within the sample.However, the frontal area length has a slight lower value in the first component and a stronger value in the second component (Table 5A, B) suggesting that the second principal component is strongly associated with variation in this character, in both PCAs.In the second analysis (Table 5B) the anterior border width has a strong negative loading on the third component, which suggests that much of the variation of the anterior border was accommodated on this axis.Besides the palpebral lobe and glabellar lengths both show positive loading on the same axis, suggesting an antagonistic pattern of growth.This could reflect a slight non size-related variability of the anterior border and a size-relationship of the palpebral lobe.
The principal components two to five are not correlated with size.Because each principal component has influence on one or more of the variables after the size effects are removed, variability patterns may be seen.Thus, the anterior border width appears to have a high influence in the third component which explains the variability in this character without a size-relationship.On the other hand, orthogonal values with different sign in the axis 2 show deformation in the sample (Hughes and Jell 1992).However, with the exception of variable 1, all variables show the same sign in axis 2 (Table 5A, B).The bivariate plots of the three principal components display the relationships among individuals (Fig. 19).The spatial distribution shows that all the specimens (inclusive of the lectotype: Fig. 19B) are in the same area of morphospace.Both species share the same morphospace and PCA does not show more than one group.The scores in the first component show differences in size, thus higher scores values correspond with larger specimens.The lower scores in the second component correspond with wider posterior glabellar width.The lower scores in third component correspond with shorter palpebral lobes whereas in the third component the frontal area width increase and the frontal area length decreases with the lower score values.
Pygidium.-ThePCA were performed with those characters of the correlation matrix (n = 99) (Table 6).The eigen-val-  7) and the first and second components account for most of the variation within the sample (97.4%).This fact suggests that the most important variation within the sample is accounted for the size in the first axis.However, some characters show slight differences such as the posterior pygidial width (transversal) and the pygidial axial length (sagittal) which has slightly lower values in the first component but higher values in the second component, suggesting that this principal component reflects variability in these characters.The anterior pygidial width (transversal) possesses a high value in the third axis which suggests that the variability is reflected in this axis.The pygidial length (sagittal) and the pygidial spine length (exsagittal) also show some variability on the third axis.This could reflect a slight non size-related fluctuation in these variables.These characters show antagonistic behavior for the posterior pygidial width (transversal) and the pygidial axial length (sagittal).Thus, these characters appear to have had high influence on the second and third axis which explains the variability in these non size-related characters.The variation patterns are assessed with bivariate plots of the three principal components (Fig. 20).Their spatial distribution shows that the specimens share the same area of morphospace.Differences in size is shown in the first and second component, high score corresponds with large specimens (Fig. 20A).The lower scores in the second com-ponent correspond with the largest pygidial axial length whereas smaller posterior pygidial width corresponds with lower scores in this axis (Fig. 20B).In the third component the anterior pygidial width correspond with an increase and the exsagittal pygidial length with a decrease of higher score values.
Discussion.-Eccaparadoxides specimens from the Purujosa trilobite assemblage show variable cranidial and pygidial morphology.Nominal and ordinal, bivariate and multivari- ate analyses show a continuous variation within the sample supporting the idea that only a single species is preserved in the sample.
The PC shows that most of this variability is due to different ratios among different sizes as it is the case in most of the trilobite species (Geyer 1990;Hughes 1994;Hughes and Chapman 1996;Sundberg 1996).However, the proportion of variance accounted for PC1 is smaller compared to many other studies of trilobites (Hughes 1994 and references herein).This suggests that there is a considerable variability accounted for PC2 and PC3.The PCA does not provide any evidence that the sample contains more than one morphotype.There is a wide overlap among all specimens in the sample.The results of the PCA are consistent with results obtained by bivariate analysis.PCA's show an antagonistic relationship between characters (e.g., the anterior border width increases whereas the palpebral lobe length decreases).The anterior border width is related with the "S" parameter described by Liñán-Guijarro (1978: 159).This parameter measures how far the facial suture is related to the anterior border width and the posterior border width, so that there are four main parameters: (i) parallel and tangent (ii) parallel and secant, (iii) convergent and tangent, and (iv) convergent and secant, in addition trilobites may show different combination of these parameters.Dies Álvarez et al (2010) described Eccaparadoxides mediterraneus as having the S parameter secant and slightly convergent.However, as a result of this variability some specimens have this parameter parallel (Figs.4G, 5F,  H) or secant (Fig. 4E, C) but even some specimens have a slight tangent parameter (Fig. 10A, B).Therefore the S parameter may not be fully diagnostic for discriminating species within the same paradoxidid genus and it needs to be applied carefully.PCA in the pygidia shows a continuous variation in the pygidial length (sagittally and exsagittally) and with antagonistic characters, the pygidial axial length and the posterior pygidial width.Liñán and Gozalo (1986) and Gozalo et al. (2003) suggested an intraspecific dimorphism with two types of pygidia, a "morphotype A" with a long pygidial axial length and short pygidial length (sagittally and exsagittally), and "morphotype B" with a short pygidial axial length and long pygidial length (sagittal and exsagittal).However, the PCA demonstrates continuous variation so that the morphotypes cannot be discriminated.A similar variability was described for the pygidia of Paradoxides davidis by Bergström and Levi-Setti (1978).
PCA allows assessing deformation effects.Hughes and Jell (1992) suggested that PC2 shows a pattern consistent with the general pattern of deformation that was affecting the entire sample.This pattern reflects the tendency for variation in some aspects of length measurements to covary with other width measures, in this sense that the lengths and widths generally vary in opposite directions.Because all the specimens come from the same bed, specimens suffered the same deformation.For that reason, if deformation has influenced the shape, we can expect to see the PC2 with variables covarying.However, the results show the same sign for lengths and widths in axis 2 (see Tables 5, 8).Nixon and Wheeler (1990: 218) suggested a species concept as "the smallest aggregation of populations [...] diagnosable by a unique combination of character states in comparable individuals".Continuous variation within an assemblage hinders subdivision of phylogenetic species (Wiley 1978).Thus, the continuity in morphology provides evidence for the integration of species.Eccaparadoxides specimens from the Purujosa trilobite assemblage show a continuous spectrum in all nominal, ordinal and metric characters.This continuity suggests that there is only a single species of Eccaparadoxides present in the studied bed at Purujosa.Because the type specimens of the two species also lie within the range of variation seen at Purujosa their status as separate species is challenged.

Conclusions
The analysis presented herein is based on a plethora of specimens, many of which are articulated, and indicates: (i) when only a small number of specimens is available, particular characters may appear to designate different species; (ii) when larger sample sizes are available the differences between the specimens are revealed to be inconsistent.For example, out of the 101 cranidia 45% show two glabellar furrows, traditionally regarded as a character of E. pradoanus (sensu Liñan and Gozalo 1986); the remaining 55% of the specimens in the sample show three glabellar furrows, which traditionally would place them under E. mediterraneus (sensu Dies Álvarez et al. 2010).Previous studies showed that the occurrence of posterolateral spines is useful to discriminate both species (two long spines E. mediterraneus; spines absent in E. pradoanus).Nevertheless, a high variability exists among small specimens as well as taphonomic sorting according to the size.Some beds bear large specimens whilst others yielded small specimens.The studied bed yielded mainly small and medium-sized specimens, however, a thorough sampling allows obtaining specimens of the full size-spectrum (Fig. 21).Therefore specimens of less than 15 mm in total length appear to have spines of random length or lack spines, those of more than 15 mm total length always have posterolateral spines.Others features, such as ornamentation, have been used as diagnostic characters in Eccaparadoxides but the ornamentation is very changeable and taphonomic processes play an important role in their preservation.
The high variability shown in most of the characters of Eccaparadoxides pradoanus suggests that other species of Eccaparadoxides, and likely also Acadoparadoxides spe-cies, have been taxonomically over split.However, although Eccaparadoxides species are not commonly used in biostratigraphy, some species have been used for biostratigraphic zonation: Eccaparadoxides benetti and E. eteminicus in Avalonia (Landing and Westrop 1996;Geyer 1998;Geyer and Landing 2001); E. insularis and E. pinus in Baltica (Westergård 1936); E. pusillus in the Czech Republic (Fatka et al. 2004); E. asturianus and E. sdzuyi in Spain (Liñán and Gozalo 1986;Álvaro and Vizcaïno 1998;Gozalo et al. 2008); and E. macrocercus in Spain and France (Josopait 1972;Courtessole et al. 1988).For that reason, morphological and biometrical studies within population and between populations should carried be out before revisions of biostratigraphic zonations, which would be affected after these types of studies.
Description.-Thelectotype is an articulated specimen with 13 thoracic segments with the thoracic rear part slightly flexed.The glabella show SO strongly curved forward and S1 strongly curved backward, S2 apparently not transglabellar (probably due to preservation), but sketched transglabellar by Verneuil and Barrande in Prado et al. (1860); S3 and S4 barely visible (Fig. 7B).The thorax is poorly preserved but shows the short horizontal inner part of each pleural segment and a poorly developed fulcrum.The paralectotype's cranidia show the same features as those from the Purujosa trilobite assemblage, one of them (Fig. 7C) only bear two transglabellar furrows (S1 and S2) and the other cranidium has two discontinuous (S3 and S4) and S4 is fade.EM 170 094 and EM 170 095 (Fig. 7E, F) show the rear part of two specimens, the pleural spines are very large and the rhachis is narrow.The pygidium of EM 170 094 (Fig. 7E) has a low spinosity-degree and the maximum width is in the vertex.The specimen EM 170 096 (Fig. 7G) is a big pygidium with the typical shape of large pygidia where maximum wide, is curved not formed an angle.
Stratigraphic and geographic range.-SouthwestSardinia (Italy), base of the Cabitza Formation, middle Cambrian, corresponding to the upper Caesaraugustan to lowermost Languedocian of the Mediterranean chronostratigraphy (Pillola et al. 2002).Iberian Chains (northeast Spain), Murero and Borobia formations.Cantabrian Mountains (northwest Spain), Oville Formation.In both areas the species is found in the Solenopleuropsis riberoi Biozone to the Solenopleuropsis thorali Biozone (upper Caesaraugustan to Languedocian; middle Cambrian).Montagne Noire, southern France: Levels B to F in the Ravin du Brian and Coulouma sections, corresponding to the upper Caesaraugustan to Languedocian of the Mediterranean chronostratigraphy (Courtessole 1973;Álvaro and Vizcaïno 1998).

Fig. 3 .
Fig. 3. Common patterns of glabellar furrows within the Purujosa trilobite assemblage, Solenopleuropsis thorali Biozone, middle Cambrian, Spain. A. Specimen with two transglabellar (S1 and S2), two discontinuous (S3 and S4), and a convergence of SO and S1.B. Specimen with two transglabellar furrows (S1 and S2), one discontinuous furrow (S3), and a convergence of SO and S1. C. Specimen with two transglabellar furrows (S1 normal and S2 shallow medially), two discontinuous (S3 and barely visible S4), and convergence of SO.D. Specimen with two transglabellar furrows (S1 normal and S2 shallow medially), two discontinuous (S3 and barely visible S4), without convergence of any furrow.E. Specimen with two straight transglabellar furrows (S1 and S2), S2 shallow medially, and a shallow S3.F. Specimen with only two straight transglabellar furrows (S1and S2).These patterns are seen in the 95% of the specimens although it is likely that other patterns can be found.Note that the ornamentation is strong in A and weaker in D and absent in E and F.

Fig. 4 .Fig. 5 .
Fig. 4. Cranidia of paradoxidid trilobite Eccaparadoxides pradoanus (Verneuil and Barrande in Prado et al., 1860) showing differently developed glabellar furrows and preservation patterns common in mudstone from the Purujosa trilobite assemblage, Solenopleuropsis thorali Biozone, middle Cambrian, Spain. A. Meraspid cranidium with one glabellar furrow and S2 barely visible, the arrow points to preglabellar field (MPZ2011/2).B. Meraspid cranidium with two glabellar furrows, the arrow points to preglabellar field, right arrow points to the exsagital extension of the anterior facial branch of facial suture which touch the side of the glabella (MPZ2011/3).C. Holaspid cranidium with two continuous glabellar furrows and one no continuous and convergence of SO and S1 (MPZ2011/4).D. Holaspid cranidium with two continuous glabellar furrow and convergence of S1 (MPZ2011/5).E. Holaspid cranidium with two continuous glabellar furrows and two barely visible discontinuous two glabellar furrow, SO and S1 almost without convergence, arrow points to the exsagital extension of the anterior facial branch of facial suture which touch the side of the glabella (MPZ2011/6).F. Holaspid cranidium with two continuous glabellar furrows and two barely discontinuous glabellar furrows; SO and S1 almost without convergence (MPZ2011/7).G. Holaspid cranidium with two glabellar furrows; SO and S1 almost without convergence, arrow points to the exsagital extension of the anterior facial branch of facial suture along the side of the glabella (MPZ2011/8).H. Holaspid cranidium with two continuous glabellar furrows and two discontinuous, SO and S1 convergent, arrow points the genal caeca (MPZ2011/9).All photographs are taken from latex casts and internal moulds covered by sublimated NH 4 Cl.

Fig. 10 .Fig. 11 .
Fig. 10.Articulated specimens of paradoxidid trilobite Eccaparadoxides pradoanus (Verneuil and Barrande in Prado et al., 1860) with a heteronomous thorax, from the Purujosa trilobite assamblege, Solenopleuropsis thorali Biozone, middle Cambrian, Spain.A, B, D. Latex casts.A. MPZ2011/23, slightly disarticulated specimen with macropleural spines, the arrow in the cephalon points to the exsagittal extension of the anterior facial branch of facial suture which touch the side of the glabella.B. MPZ2011/24, specimen with large macropleural spine, the arrow in the cephalon points to the exsagittal extension of the anterior facial branch of facial suture and in the trunk points the macroplural spine.D. MPZ 2011/72, immature specimen with the rear part of the thorax fold, the arrow in the trunk points the macroplural spine.C, E. Internal moulds.C. MPZ2011/25, specimen with large macropleural spine, the arrow in the cephalon points to the exsagittal extension of the anterior facial branch of facial suture which touch the side of the glabella, the arrow in the trunk points the macroplural spine.E. MPZ2011/26, large specimen with two large macropleural spines.Note that these macropleural spines are relatively shorter than in the specimens shown in B or D. All specimens are covered by sublimated NH 4 Cl.Scale bars 5 mm.→
(McNamara 1986)some holaspids retain the feature from the meraspid stage for a period of holaspid growth may explain this fact, which is called post-displacement paedomorphosis(McNamara 1986).However, further studies with more articulated specimens of different sizes and development of the macro and microspinous conditions are necessary to assess this hypothesis.Nevertheless, multiple holaspids with either the heteronomous or the homonomous condition have only been reported from collections of the Iberian Chains, while holaspid specimens from the Cantabrian Mountains (North Spain), Montagne Noire (South France) and from Sardinia bear only homonomous thoraces.This suggests either a flexible developmental response of some individuals to local environmental conditions, or interpopulational variation.Both possibilities imply that Eccaparadoxides may have had a flexible developmental genome, rather than a series of genetically canalized polymorphs.Number of thoracic segments.-Eccaparadoxidespradoanusbears17segmentsaccording toSdzuy (1961: 605), whereas E. mediterraneus bears 18 segments according to DiesÁlvarez et al. (2010: 103).However, both putative species of Eccaparadoxides vary in the number of thoracic segments within the Purujosa sample, each having representatives with 16, 17, or 18 thoracic segments

Table 1 .
Reduced major axes for each cranidial linear measure.

Table 2 .
Reduced major axes for each pygidial linear measure.

Table 4 .
Eigenvalues and percentage of variance for cranidial analyses.

Table 8 .
Relationship of each variable to each principal component is shown by the eigenweight in pygidial analysis.

Table 7 .
Eigenvalues and percentage of variance accounted for by each principal component in pygidial analysis.