A quantitative evaluation of evolutionary patterns in opercle bone shape in Saurichthys (Actinopterygii: Saurichthyidae)



Despite an impressive radiation of more than 30 species in the wake of the end-Permian mass extinction, the taxonomic study of Saurichthys has suffered from a lack of universally diagnostic features and a lack of tested quantitative schemes that can be applied to analyse interspecific morphological differences. In this study, we provide an initial quantitative framework for morphological evolution in Saurichthys by focusing on a single bone, the opercle and exploring patterns of interspecific variability in shape using outline-based geometric morphometrics and linear measurements. For the six species examined, comprising 155 specimens and representatives from the Early, Middle and Late Triassic, our results indicate that interspecific shape differences largely reflect an anterior–posterior dimension decrease (= craniocaudal direction) as the dorso-ventral dimension remains similar. In contrast, intraspecific variability in shape is subtle and spread across the outline of the bone, such that counter-acting dimension differences (increase/decrease) were found to occur along a single margin at oblique axes in several species. Our quantitative scheme, which is widely applicable to other groups, provides a useful description of the broad modes of opercle shape change that may help as a starting framework from which to develop character states for opercle morphology in future study.

Understanding the patterns and processes underlying morphological diversification is a central theme of evolutionary research. To reconstruct the evolutionary history of a clade, the characterization of morphological diversity and its relation to taxonomy and systematics is a fundamental step, especially when considering fossil taxa. In this article, we examine the evolution of shape in a skull bone, the opercle, in an extinct fish clade. The opercle is typically well preserved, its variation can be of taxonomic relevance (e.g. Hilton 2003; Hilton et al. 2011; López-Arbarello 2012; Marceniuk et al. 2012), and several works have used the bone as a model to examine the relation between morphological evolution and development (e.g. Kimmel et al. 2005).

Saurichthys was a highly diverse and nearly globally distributed genus of actinopterygian fishes occurring in the whole Triassic, occupying both marine and freshwater realms (e.g. Beltan and Tintori 1980; Rieppel 1985; Kogan et al. 2009; Wu et al. 2009). Despite an impressive diversity of more than 30 species in the wake of the end-Permian mass extinction, the taxonomic study of Saurichthys has suffered from a lack of universally diagnostic features such that new species have been created from isolated tooth and jaw fragments (e.g. 1833–43; Schmid 1861); a lack of consensus regarding the mode of character evolution and hence difficulty in tracing patterns of morphological change (Schmid and Sánchez-Villagra 2010); and a lack of tested quantitative schemes that can be applied to analyse interspecific morphological differences.

Saurichthyidae (sensu Stensiö 1925), most recently placed together with Birgeria near the base of Acipenseriformes (Gardiner and Schaeffer 1989; Gardiner et al. 2005), were the first fishes to have developed an elongated, slender body with posteriorly placed pelvic, anal and dorsal fins in addition to a pointed, oblong-shaped slender snout and long jaws (Mutter et al. 2008; Romano et al. 2012). Saurichthys is reconstructed as bearing physical resemblance to the modern day garfish (Belone belone), and the slender body and snout suggest a lifestyle as a fast-swimming predator. The readily distinguishable morphology of Saurichthys has resulted in a relatively well-documented record of its palaeogeography (see Romano et al. 2012 for review), but many of the recovered remains are unnamed or only briefly described under ‘Saurichthys sp.’ (Mutter et al. 2008), making difficult the assessment of morphological transformations during the Triassic and emphasizing the necessity of a taxonomic revision for the genus.

The aim of this study is to provide a first step towards a quantitative framework for morphological evolution in Saurichthys by focusing on a single bone, the opercle and exploring patterns of interspecific variability in shape and their potential relevance for aiding taxonomic study. The opercle bone has been the subject of previous quantitative studies in extant fish (e.g. Arif et al. 2009; Kimmel et al. 2005, 2008, 2012a) because it is well suited to geometric morphometric description and the genetic underpinnings of its development and growth have been well documented in zebrafish (e.g. Kimmel et al. 2010). The opercle is a large, flat craniofacial bone that articulates at its anterior margin with the preoperculum and supports the gill cover in bony fish, therefore playing an important functional role in the mechanism of gill ventilation (Hughes 1960). Arising early in development (Cubbage and Mabee 1996) and growing appositionally, the bone undergoes major transitions in shape as a consequence of differential outgrowth during ontogeny (Kimmel et al. 2010). Recently, attention has drawn towards understanding the microevolutionary dynamics of changes in opercle shape, using sticklebacks as a model system. Kimmel and colleagues have published a series of landmark-based geometric morphometric studies showing that opercle morphology differs between oceanic and freshwater stickleback populations, and that the evolutionary axes of shape change have evolved in parallel among geographically distant populations on a rapid timescale (Kimmel et al. 2008, 2012a, 2012b).

The trends documented so far for stickleback populations provide a platform from which to base a quantitative approach to examining opercle shape evolution in the fossil record. The operculum is typically conspicuous among craniofacial bones and therefore quite easy to identify in fossil material, and it is often preserved in lateral orientation, thereby permitting an accurate estimation of its flattened shape to be captured by photograph. In contrast to the landmark-based approach of previous microevolutionary studies (e.g. Kimmel et al. 2008), here we adopt an outline-based approach to permit the quantification of subtle differences in opercle shape across six species of Saurichthys, comprising representatives from the Early, Middle and Late Triassic from deposits in Madagascar, Italy and Switzerland.

Institutional Abbreviations

GBA, Geological Survey of Austria, Vienna, Austria; MCSN, Museo cantonale di storia naturale, Lugano, Switzerland; MNHN, Muséum national d'Histoire naturelle, Paris, France; PIMUZ, Palaeontological Institute and Museum, University of Zürich, Switzerland.

Material and Methods

Data set

We examined 155 opercles belonging to specimens of Saurichthys (see Fig. 1 for examples). Only adult specimens were included and assessed based on skeletal measurements documented in the works of Rieppel (1980, 1985, 1992). The sample comprised identified material from specimens of Saurichthys costasquamosus (N = 4), Saurichthys curionii (N = 18), Saurichthys madagascariensis (N = 16), Saurichthys macrocephalus (N = 3), Saurichthys paucitrichus (N = 1) and Saurichthys striolatus (N = 28) (Table 1). These species are well diagnosed morphologically (1944-45; Lehman 1952; Griffith 1959; Rieppel 1980, 1985, 1992). Additionally, we included opercles belonging to unidentified specimens from stratigraphically controlled Middle Triassic sections in Switzerland, located at Monte San Giorgio (Ticino) and at the Ducan mountains, near Davos (Graubünden), referred to Saurichthys sp. indet., and hereafter ‘unknown’ (Table 1, Table S1; Furrer 1995, 2009; Bürgin 1999; Furrer et al. 2008; Stockar 2010; Stockar et al. 2012). Four juvenile specimens, comprising two specimens of S. curionii (PIMUZ T 4451, and T 4452) and two specimens of S. costasquamosus (PIMUZ T 309 and T 603), were also examined qualitatively for ontogenetic comparison. Because our study has focused only on well-diagnosed species, it should be noted that our temporal sampling of Saurichthys diversity is unequally distributed across the records of presently known occurrences for the Triassic (Romano et al. 2012). We sample four species from the Middle Triassic record at an interval (Anisian–Ladinian) in which the majority of specimens identified to species are currently known (see Romano et al. 2012). Our sampling from the Early Triassic (only S. madagascariensis) does not include several species from the Smithian of Spitsbergen because their diagnostic features are a matter of debate and require a complete taxonomic revision (Mutter et al. 2008; Romano et al. 2012). Compared to the Early and Middle Triassic, diversity of Saurichthys is comparatively reduced in the Late Triassic (Romano et al. 2012), and in this interval, we also sample only a single species (S. striolatus), occurring in the stage comprising the most species diversity (Carnian) presently known from the Late Triassic (Romano et al. 2012). The studied species differ in size: S. striolatus is considerably smaller (100–181 mm, Kner 1866) than the other species examined, all of which typically exceed 250 mm in total body length, the exception being S. paucitrichus, which is slightly smaller (210–260 mm, Rieppel 1992). Raw size values for the opercle of each species (Table 1) indicate a range of opercle dimensions from the smallest species S. striolatus (4.97 mm OP height) to the largest, S. macrocephalus (34.24 mm OP height). Specimens examined in this study are housed in GBA, MCSN, MNHN and PIMUZ.

Table 1. General overview of specimens included in this study. Refer to Table S1 for further details including specimen identifier, chronology and measurement scheme
SpeciesNAverage opercle height and range (mm)Average opercle length and range (mm)Age: Early (E)/Middle (M)/Late (L) TriassicLithostratigraphy
Saurichthys striolatus 284.87 (3.32–6.67)4.90 (3.36–6.81)LRaibl beds
Saurichthys paucitrichus 115.3710.74MBesano Fm.
Saurichthys costasquamosus 410.73 (8.12–12.15)5.48 (4.96–5.78)MBesano Fm.
Saurichthys macrocephalus 334.24 (16.91–47.26)19.63 (10.68–25.98)MMeride Fm.
Saurichthys curionii 1820.78 (12.05–32.87)15.38 (10.29–25.32)MMeride Fm.
Saurichthys madagascariensis 1621.18 (18.08–25.79)16.95 (13.09–23.08)ESakamena moyenne
Unknown7919.17 (9.04–37.09)13.41 (5.78–26.4)MMeride Fm., Besano Fm.
Figure 1.

Examples of opercles: A, Saurichthys madagascariensis (MAE481ab). B, Saurichthys costasquamosus (PIMUZ T 4102). C, Saurichthys curionii (PIMUZ T 4104). D, Saurichthys sp. (PIMUZ T 5736). Scale bars represent 10 mm.

In all cases, specimens were carefully selected for morphometric analyses such that each operculum was not or only minimally damaged at the edges of the bone and preserved in a lateral orientation, enabling the outline shape of the bone to be adequately captured by a photograph.

Each specimen was photographed in a standardized manner: a digital camera mounted on a tripod was used to capture an uncompressed image of the operculum in lateral orientation with a scale bar placed on the same plane as the bone. For some images, in which the operculum was not preserved in life position, subsequent processing was undertaken in Adobe Photoshop to re-orient the bone to exhibit a left-side view, as shown in Figure 2.

Figure 2.

Linear and geometric morphometric measurement schemes used in this study. Linear measurements comprise operculum length (OP length) and operculum height (OP height). Outline-based geometric morphometric methods were used to capture the entire outline of the bone using 100 equi-distant landmarks (open circles). A spatially homologous point (filled circle) was defined as starting point for each specimen.

Because some opercles were isolated or only the left or right bone was available to us, and shape analyses are sensitive to subtle differences in morphology, before beginning data collection, we initially created a subset of samples that had both the left and right opercle preserved (N = 10 specimens, N = 20 opercles). To assess the difference in shape between left and right bones, we collected and analysed morphometric data for the subset as described below. In all cases, interspecimen shape differences were much larger than intra-specimen differences between left and right opercles, as indicated by anova results of geometric morphometric shape data (within specimen sum of squares = 0.003–0.010; between specimen sum of squares = 0.216–0.279). Therefore, we proceeded to combine measurements of left and right bones in the final data set.

Morphometric analyses

To compare opercle size and shape, two sets of different measurements were collected for each specimen. First, a morphometric quantitative scheme previously created by Rieppel (1985) was used to assess simple linear measurements of height and length for each opercle (OP height, OP length: Fig. 2). The measurements capture the longest axis in the dorsoventral direction (OP height) and the craniocaudal direction (OP length). We subdivided the two measures and report proportion values for the OP height/OP length calculation. Second, geometric morphometric methods were used to identify landmarks on photographs of each opercle (e.g. Fig. 3). Geometric morphometric methods are a useful toolkit for quantifying morphological shape, producing data that are easily visualized in morphospaces and may be further analysed with multivariate statistical methods (e.g. Bookstein 1991; Adams et al. 2004; Mitteroecker and Gunz 2009). An outline-based approach was chosen because the curved nature of the operculum makes difficult the identification of a sufficient number of biologically meaningful, homologous, landmark points required for an accurate description of its shape. Eigenshape analysis is based upon the definition of additional points of reference, or so-called semilandmarks (MacLeod 1999) that are used to fill landmark-depleted regions, and in doing so enable the shape difference located in-between landmarks to be sampled, and the global aspect of a boundary outline to be evaluated (Wilson et al. 2011). Eigenshape analysis has proven to be successful in elucidating subtle shape variation in a wide variety of contexts (e.g. Polly 2003; Krieger et al. 2007; Wilson et al. 2008; Astrop 2011) and is particularly suitable for this study because it affords the possibility to examine localized variation in opercular shape.

Figure 3.

Lateral view of a head of Saurichthys macrocephalus (PIMUZ T 4106) from the Cassina beds (Early Ladinian) at the Monte San Giorgio locality in Switzerland, showing the opercle bone (op). Scale bar represents 20 mm.

For each opercle, the outline of the bone was traced using the software tpsDig (v. 2.16, Rohlf 2010) (Fig. 2). A type II (Bookstein 1991) landmark was defined as the starting point for each outline and is described as the maxima of curvature on the dorsal margin of the bone (Fig. 2). The dorsal margin was chosen rather than the anterior-most projection of the operculum at its point of articulation with the preoperculum, because in some cases the latter point was difficult to unambiguously identify on some specimens and was therefore considered more subjective and susceptible to replication error. Each outline was re-sampled to create 100 equi-distant landmark points. Cartesian x–y coordinates of these landmark points were converted into the phi (φ) form of the Zahn and Roskies (1972) shape function, required for eigenshape analysis (MacLeod 1999). Eigenshape analysis was performed using FORTRAN routines written by Norman MacLeod (NHM London). The method is based on a singular value decomposition of pairwise covariances calculated between individual shape functions. Similar to principal component (PC) analysis, which also uses an eigen decomposition, it produces a series of mutually orthogonal latent shape vectors that represent successive smaller proportions of overall shape variation such that the greatest amount of shape variation is represented on the fewest independent shape axes. Each specimen has a series of eigenscores, representing its location along each axis, and therefore, specimens can be projected into a multi-dimensional morphospace to visualize shape differences. Our plots for ES axes can therefore be considered analogous to other studies reporting PC axes.

To provide a useful, visual schematic of the main axes of variation in opercle shape for each species, the data set was subdivided. Excluding S. paucitrichus, which is represented by only one specimen, data sets were created for each species and landmark data were analysed separately, using the eigenshape method detailed above, to generate a meanshape per species and create shape models representing the major axes of intraspecific variation in shape.

Interspecific shape differences

Canonical variates analysis (CVA) was performed on the eigenscores for all axes that in total represented 95 per cent of shape variation in the sample. Groups were defined a priori for each of the six identified species in the sample. Saurichthys specimens that had not been assigned to a species (unknown) were included as unknowns in the classification phase to determine the probabilities of their belonging to any of the a priori groupings. The ability of the canonical variates to assign a specimen to a given group was further assessed by cross-validation, in which individuals are removed from the data set prior to calculation of the canonical function and are classified based upon results generated from the remaining specimens. Differences between species groups were examined using univariate analysis of variance (anova) coupled with post hoc tests to assess whether different species showed significantly different eigenscores.

Linear measurement repeatability and precision of outline data collection

To minimize measurement error associated with the linear measurements, the length and width measurement values used for analysis were based on the average of two replicates for each specimen. To assess error measurement in relation to capturing the ‘true’ outline of the opercle, error associated with the shape variables derived from outline data sets was calculated following the method of Arnqvist and Martensson (1998). Landmark data were collected twice for a subset of 20 specimens, and outlines were interpolated for the error repeats and added to the original data set. Eigenshape analysis was used to obtain shape variables, and a one-way anova was then performed on the outputted shape variables to detect whether the among-individual variance was greater than the within-individual (repeated) variance. The repeatability (R) value scales between 0 and 1. An R value of 0 would represent a sample in which all variance is found within individuals, whilst an R value of 1 would indicate all the variance is due to differences between individuals (see Wilson et al. 2011).


Precision of outline data collection

Repeatability (R) values were calculated for eigenshape (ES) axes that represented 90 per cent of shape variance in the sample (ES1-ES6, Table S2). Overall, R values were high and ranged from 0.962 for ES1 to 0.886 for ES6. Axes accounting for smaller portions of shape variance had lower R values. This result was expected because eigen analysis will tend to recover true structure from the earlier axes, leaving the later ones containing larger portions of error as they explain successively smaller proportions of variance (e.g. Wilson et al. 2011; Wilson 2012a).

Linear measurements of opercle dimensions

Raw linear measurements were plotted for all specimens (Fig. 4) to provide a clear indication of size, because the proportions for OP height/OP length make it difficult to assess the dimension changes underlying those numbers. S. striolatus is clearly much smaller than the other species, which exhibit much overlap in the size range of c. 10–15 mm for opercle length and 15–25 mm for opercle height. Opercle height and length measurements were subdivided by one another for each specimen, and descriptive statistics were graphed for these proportions for each species group (Fig. 5). S. striolatus specimens had the smallest absolute values (average = 0.99), followed by broadly similar and greater values for S. madagascariensis (average = 1.26), S. curionii (average = 1.30) and S. paucitrichus (1.43). Lastly, the largest values belonged to specimens of S. macrocephalus (average 1.71) and S. costasquamosus (average = 1.87). anova results, based on specimens identified to species only and excluding S. paucitrichus (N = 1), revealed significant differences among species groups for OP height/OP length values (F59,4 = 73.71, p < 0.001). Opercle height measures ranged from an average of 4.87 mm for S. striolatus to 34.24 mm for S. macrocephalus, whereas average opercle length varied from 4.90 mm for S. striolatus to 19.63 mm for S. macrocephalus.

Figure 4.

Plot of raw linear measurements for opercle height vs opercle length for all specimens.

Figure 5.

Box plot of variability in values for opercle height/opercle length for six species of Saurichthys. Species’ meanshapes are illustrated above their corresponding box, and the phylogenetic relationship between species is illustrated based on Rieppel (1992).

Trends in opercle outline shape change

The first three eigenshape (ES) axes accounted for 75.8 per cent of variance in opercle shape (Fig. 6). The first eigenshape axis (ES1, 52.4 per cent variance) separated specimens of S. striolatus, having the greatest negative scores along the axis (average = −0.21), from specimens of S. madagascariensis that had ES1 scores averaging −0.043 and specimens of S. costasquamosus (average = 0.098) and S. macrocephalus (average = 0.10). Shape change along ES1, from negative to positive scores, reflected a decrease in dimension on a horizontal plane (= craniocaudal direction) of the anterior and posterior margins of the opercle coupled with perpendicular increase in dimension of the dorsal and ventral edges, as can be visualized by the vectors of displacement from the meanshape and the axes labelled ‘a’ and ‘c’ shown in Figure 6. Shape variance explained by ES2 (13.9 per cent) did not account for clear differences between species groups (Fig. 6A). Specimens of S. madagascariensis had typically lower scores along ES2, although much spread was evident in morphospace occupation, relating to extension along an oblique axis running from the posterior dorsal margin to the anterior ventral margin, coupled with dimension decrease in a perpendicular direction (Fig. 6, axes b and d). These shape changes resulted in an opercle with a highly asymmetric shape due to an extension of the upper portion of the posterior margin creating a more triangular outline. Similar to ES2, all species groups had broadly overlapping ranges along ES3 (Fig. 6B), which accounted for 9.5 per cent shape variance in the sample. Subtle variation along three oblique axes was evident in the vector displacement models for ES3 (Fig. 6B). Shape variation was quite similar in vector displacement direction to that along ES2 but ventrally offset (compare model for high ES2 score with model for high ES3 score, Fig. 6).

Figure 6.

Ordinations of eigenshape (ES) axes. Left (A), ES1 vs ES2, and right (B), ES1 vs ES3 representing successively smaller portions of variation in opercle shape for six species of Saurichthys and unidentified material from the Middle Triassic of Switzerland. Meanshape models are deformed with vector displacement lines to indicate the main axes and relative magnitude of shape change in the sample. Meanshape is located at (0, 0) in eigenshape space. The schematic (top left) illustrates the main axes of shape change in the opercle outline for ES1 (solid line, axes a and c) and ES2 (dashed line, axes b and d).

anova results indicated significant differences between species groups on ES1 (F58,4 = 56.92) and ES2 (F58,4 = 7.492), and these largely reflected differences in shape between S. striolatus and all other species (p < 0.001, Bonferroni corrected), and S. madagascariensis and all other species (p = 0.007 to <0001, Bonferroni corrected).

Intraspecific variability in opercle shape

Patterns of intraspecific shape variation differed among the five species analysed (Fig. 7). With the exception of S. striolatus (Fig. 7A) and to some extent S. curionii, the main axes of shape change did not simply mirror the interspecific patterns, but revealed additional, subtle changes along the margin of the operculum. In the case of S. striolatus, intraspecific variation along ES1 was mainly explained by a dimension increase and decrease along the anterior–posterior and dorsoventral axes, although these axes were offset slightly compared with those detected for interspecific models. A similar trend was found for S. curionii, although along the anterior margin, the displacement vectors from the upper and lower portions converged acting to create a more (negative ES1 score) or less (positive ES1 score) pronounced triangular shape of the anterior margin (Fig. 7D). Shape change models for specimens of S. madagascariensis (Fig. 7E) indicated an increase in dimension of the upper anterior margin coupled with a decrease in dimension of the lower anterior margin at negative ES1 scores and the opposite at positive ES1 scores. A similar pattern of increase and decrease in dimension occurring along upper and lower portions of one margin is also evident for S. costasquamosus along the dorsal margin (Fig. 7B) and for S. macrocephalus along the anterior margin (Fig. 7C).

Figure 7.

Meanshape models of intraspecific variation in opercle shape for Saurichthys. Models illustrated represent variation at −0.3 and 0.3 along the first eigenshape axis (ES1) for A, Saurichthys striolatus, B, Saurichthys costasquamosus, C, Saurichthys macrocephalus, and D, Saurichthys curionii, E, Saurichthys madagascariensis. The vertical line denotes the coordinate position 0,0 in eigenshape space, representing the meanshape models illustrated in Figure 3.

Canonical variates analysis

Canonical variates analysis of the eigenscores for axes ES1–ES8, accounting for 96.3 per cent of the sample variance, produced two canonical variates with significant discriminatory power, accounting for 97.2 per cent of the variance (Table 2). The first canonical variate (CV1) accounted for a large portion of sample variance (88.1 per cent) and separated specimens of S. striolatus having negative scores, compared with specimens of S. costasquamosus, S. macrocephalus and S. curionii having positive scores, associated with progressive dorsoventral increase in dimension of the opercle coupled with proximodistal decrease in dimension (Fig. 8). CV2 accounted for 9.1 per cent of variance; S. madagascariensis specimens were incompletely separated from the other species by having typically low CV2 scores, associated with an asymmetrical extension of the distal end of the opercle (Table 2; Fig. 8). It is noteworthy to acknowledge that, due to the low sample sizes for several groups, and the unequal distribution of specimens per group, the results of the classification phase are less robust than could be achieved for studies of extant taxa, which are typically not hampered by the limited availability of well-preserved fossil material. For completeness, we therefore tabulate CVA results computed with priors weighted using group-size- and equal-weighted priors. The classification phase of the CVA was used to assess the capability of the canonical model to discriminate between species groups. Results indicated 79.0 per cent of original grouped cases (87.1 per cent weighted priors) and 71.0 per cent of cross-validated grouped cases (79.0 per cent weighted priors) were correctly classified (Table 3). Unidentified specimens (unknown) were assigned to S. curionii (N = 42), S. macrocephalus (N = 18), S. costasquamosus (N = 14), S. madagascariensis (N = 4) and S. paucitrichus (N = 1) (Table 3). Because detailed information is available for the locality of each specimen and all unidentified material came from Middle Triassic Swiss localities (see Table S1), it is possible to assume that of the 79 unidentified specimens used in the CVA, those predicted to belong to S. madagascariensis (N = 4), which is believed to be an Early Triassic species, were incorrectly assigned. Further evaluation of the classification success for the unidentified specimens may be gained by considering whether the species group assignment is in agreement with known occurrences of the species at a given locality. This line of reasoning may only provide a rough approximation for specimens from Monte San Giorgio and is chosen rather than strict stratigraphic comparison because the four identified species (S. curionii, S. macrocephalus, S. costasquamosus and S. paucitrichus) are found in deposits of the same age in some cases (see Table S1). In the Middle Triassic deposits of Monte San Giorgio in Switzerland, S. curionii and S. macrocephalus are known from the Cassina beds, whereas S. costasquamosus and S. paucitrichus are found in the Besano formation (Rieppel 1985, 1992). Overall, 76.4 per cent of unidentified specimens from Cassina beds were identified as either S. curionii or S. macrocephalus, but only three of 22 unidentified specimens from beds in the Besano formation were identified as either S. costasquamosus or S. paucitrichus (Table S3). This discrepancy, however, likely reflects the smaller sample size for these two species, and whilst retained in the data set for completeness, the predictive power of the S. paucitrichus group, which had a learning sample of only 1 specimen, is effectively zero and therefore we cannot propose to be testing that any of the unknown specimens are S. paucitrichus.

Table 2. Summary results for canonical variates analysis of opercle shape (outline-based geometric mophometric methods) and opercle shape plus size (linear measurements) in Saurichthys
Canonical variateOutline dataOutline and measurement data
Percentage variance88.19.181.311.34.1
Percentage cumulative variance88.197.281.392.696.7
Canonical correlation0.9340.6430.9570.7770.598
Wilks' lambda0.0620.4820.0140.1720.433
Table 3. Canonical variates analysis classification results based on opercle shape data. Number of original group cases and number of cross-validated group case (denoted in parentheses). Percentage classification success is for cross-validated result
  Saurichthys striolatus Saurichthys paucitrichus Saurichthys costasquamosus Saurichthys macrocephalus Saurichthys curionii Saurichthys madagascariensis Classification success (per cent)N
Computed from equal priors
 S. striolatus 23 (22) 0 (0)0 (0)0 (0)0 (0)0 (1)95.723
 S. paucitrichus 0 (0) 1 (0) 0 (1)0 (0)0 (0)0 (0)01
 S. costasquamosus 0 (0)1 (1) 2 (1) 0 (1)1 (1)0 (0)25.04
 S. macrocephalus 0 (0)0 (0)0 (1) 2 (1) 1 (1)0 (0)33.33
 S. curionii 0 (0)1 (2)3 (2)3 (4) 11 (10) 0 (0)55.618
 S. madagascariensis 0 (0)2 (2)0 (1)0 (0)1 (1) 10 (9) 69.213
Computed from group size–weighted priors
 S. striolatus 23 (23)0 (0)0 (0)0 (0)0 (0)0 (0)100.023
 S. paucitrichus 0 (0)1 (0)0 (0)0 (0)0 (0)0 (1)01
 S. costasquamosus 0 (0)0 (1)1 (0)0 (0)3 (3)0 (0)04
 S. macrocephalus 0 (0)0 (0)0 (1)2 (0)1 (2)0 (0)03
 S. curionii 0 (0)0 (0)0 (1)1 (1)17 (16)0 (0)88.918
 S. madagascariensis 0 (0)0 (0)0 (0)0 (0)3 (3)10 (10)76.913
Figure 8.

Ordination of the first (CV1) and second (CV2) canonical variates based on eigenscores (shape descriptors) of opercle shape variation in a sample of six species of Saurichthys. CV1 and CV2 account for 97.3 per cent of the sample variance (see Table 2). Unidentified material (unknown) from the Middle Triassic of Switzerland was included in the classification phase.

A second CVA was computed in which linear measurement (= size) and geometric morphometric (= shape) data were combined. Three canonical variates accounted for 96.7 per cent of the sample variance (Table 2). Similar to the CVA using only shape data, CV1 (81.3 per cent variance) separated specimens of S. striolatus having negative scores, compared with specimens of S. costasquamosus, S. macrocephalus and S. curionii having positive scores, indicating the main axis of interspecific variation is explained by size-related shape change (Fig. 8). The classification phase was slightly more successful due to including opercle height and length measurements and yielded a classification success of 90.3 per cent for original grouped cases and 85.5 per cent for cross-validated cases. The additional of opercle height and length measurements to the CVA resulted in a greater proportion of all unidentified specimens being assigned to S. curionii (N = 56 compared with N = 43) and S. costasquamosus (N = 16 compared with N = 14) and comparatively fewer to S. macrocephalus (N = 2 compared with N = 18). Using the combined data set, a broadly similar number of unidentified specimens from the Cassina beds were assigned to either S. curionii or S. macrocephalus (74.5 per cent), and for the Besano formation, an additional two unidentified specimens were assigned to S. costasquamosus or S. paucitrichus.


In this study, we present the first geometric quantification of opercular morphology in any Mesozoic fish. Opercular size and shape patterns in the species of Saurichthys examined here can be organized into three broad morphotype groupings. Forming one group, the opercles of S. macrocephalus and S. costasquamosus are characterized by an only slightly curved distal margin, an almost uncurved anterior margin (Fig. 6) and a relatively high value for OP height/OP length (average 1.71–1.87) indicative of a tall, narrow bone (Fig. 5). The opercles of S. madagascariensis, S. curionii and S. paucitrichus can be loosely collected into a second group, typically having lower OP height/OP length values (average 1.26–1.43) and a greater extension and curvature of the posterior margin without extension of the dorsoventral axis (meanshapes, Fig. 7). Forming a third group, the opercle of S. striolatus has a much smaller OP height/OP length value (average 0.99) and a greater increase in dimension along the posterior margin coupled with considerable dorsoventral dimension decrease to result in a bone that is anteroventerally almost elliptical (meanshapes, Fig. 6). These groups provide a useful indication of the broad modes of opercle shape change and may help to provide a starting point from which to develop character states for opercle morphology in future study.

Subtle interspecific differences in opercle shape resulted in a reasonably high level of classification success for the CVA (79.0 per cent), suggesting opercle shape provides a useful tool to consider in addition to other species-specific diagnostic features when assessing unassigned material. When shape descriptors are combined with linear measurement data, the success of classification increases to 90.03 per cent, underlining the interaction between size and shape and size-related shape changes, as further evidenced by species clustering for raw linear measurements of OP height and OP length. In both analyses, the first canonical variate (CV1) explained a large proportion of shape and size variation (81.3–88.1 per cent), and this could be attributed to counteracting increase and decrease in dimension along the proximodistal and dorsoventral axes of the opercle, as illustrated in eigenshape models along ES1 (Fig. 6). Although our classification results are sample dependent and may not be widely applicable across other species of Saurichthys not sampled in this study, our methods are broadly applicable to other groups and the shape changes we uncover here may likely have a functional base, which would be worthwhile to explore further for palaeobiological reconstructions of Saurichthys. Comparative study of gill ventilation in several groups of marine teleosts has shown differential importance of the buccal (pressure) and opercle (suction) pumps to respiration cycles across different species, depending upon the habit of the fish (Hughes 1960), and the large diversity in both opercle shape and size among teleosts is perhaps not surprising in this context and likely adaptive (Kimmel et al. 2005). For instance, the adaptive value of the opercle pump is clear for benthic fish, in which a current is drawn across the gills during a relatively long part of the respiratory cycle to counteract almost stationary bottom waters (Hughes 1960). Also, fish feeding on benthic prey typically use a suction feeding mechanism, whereas in contrast fish feeding on planktonic prey rely on ram feeding (Gerking 1994; Willacker et al. 2010), and these differences in employed feeding mechanism are known to be reflected in body shape and bone morphology among benthic and limnetic morphotypes in cichlids (e.g. Barluenga et al. 2006; Clabaut et al. 2007). Eigenshape analysis results also detected further, more subtle shape differences along oblique axes of the opercle (Fig. 6), although these did not readily discriminate between species and encapsulated a relatively broad range of realizable shape change along those dimensions, as indicated by the spread of points along ES2 and ES3 (Fig. 6). Similar to a recent attempt to use a morphometric scheme to classify skulls of Saurichthys elongatus, Saurichthys ornatus and Saurichthys wimani (Romano et al. 2012), we also note a considerable amount of intraspecific variability, and in Figure 7 we provide a visualization of the main modes of intraspecific variation in opercle shape for five of the six examined species. Intraspecific modes of shape variation differ between species and the vectors of displacement from the meanshape indicate that an increase and decrease in dimension can also occur along upper and lower portions of one margin of the bone, as is the case for example in S. madagascariensis, S. costasquamosus and S. macrocephalus.

At present, a robust phylogenetic framework for Saurichthys is lacking, and the matrix constructed by Rieppel (1992), containing eight characters and nine species, represents the most comprehensive current analysis. The position of Saurichthys in actionopterygian phylogeny is also not completely resolved, although most studies place it together with Birgeria near the base of Acipenseriformes (e.g. Gardiner and Schaeffer 1989; Rieppel 1992; Gardiner et al. 2005). A number of trends in morphology are recognizable over the evolutionary history of Saurichthys (Rieppel 1985, 1992; Mutter et al. 2008), and recently Schmid and Sánchez-Villagra (2010) inferred that, based on evidence from patterns in extant fish mutants, a change in molecular signalling pathway was a likely initial step underpinning diversification in the genus. Anatomical changes in the evolution of Saurichthys can largely be summarized as a process of gradual loss in number of scales, in number of fin rays and in number of dermal elements of the skull and also a shortening of the postorbital region (Rieppel 1985; Schmid and Sánchez-Villagra 2010; Romano et al. 2012). In the case of the latter two trends, the apparent net effect on the shape of the cranium is an anterior–posterior shortening and slightly increased posterior height (Romano et al. 2012). The opercle meanshape models and OP height/OP length metrics plotted on Saurichthys phylogeny in Figure 5 reveal an increase in opercular height at the expense of length from the Early Triassic (S. madagascariensis) to the Middle Triassic (S. costasquamosus) followed by a reversal in S. striolatus, occurring in the Late Triassic, to an opercle shape that is more similar to S. madagascariensis in which greater length is achieved at the expense of height. The shape of the opercle for S. striolatus differs from that of S. madagascariensis by being less tapered towards the posterior margin, such that the overall shape appears less triangle like because the height at the posterior margin is more similar to that at the anterior margin. The shape of the operculum in S. madagascariensis (Rieppel 1980) and also in Spitsbergen saurichthyids (Stensiö 1925) is relatively large and wide. This feature has previously been speculated to reflect a fusion of the operculum with an originally distinct subopercular element and possibly also an interopercular element. Using average values of ES1 and ES2 scores computed for each species, we also projected opercle shape into a phylomorphospace (Fig. S1, e.g. Sidlauskas 2008; Meloro and Jones 2012) to visualize the occupation of shape space in relation to time (Early, Middle and Late Triassic), and this plot further emphasizes the aforementioned trends. These are general trends, currently dependent upon the limited phylogenetic framework available for Saurichthys, and therefore, we do not suggest any polarity of these shape changes without more rigorous testing across a wider range of species.

The genetic basis for the development of the opercle is known from detailed studies on zebrafish opercle cell morphogenesis (e.g. Miller et al. 2007; Talbot et al. 2010), and recent work by Kimmel and colleagues has documented microevolutionary patterns of opercle change in stickleback populations, using both genetic and geometric morphometric approaches (Kimmel et al. 2005, 2008, 2012a, 2012b). In sticklebacks, repeated patterns of shape evolution on a relatively short timescale have been shown as a response to changing ecological conditions, specifically the colonization of freshwater habits by ancestral marine populations (Arif et al. 2009; Kimmel et al. 2012a). Sticklebacks are only very distant relatives to Saurichthys, and therefore a direct comparison would be inappropriate, nevertheless the general evident trend of plasticity in opercle morphology is also reflected in our data, and the principal axes of shape variation are the same: an anterior–posterior dimension increase and a dorsoventral dimension decrease (Kimmel et al. 2008, 2012a). All of the species we examine here are known only from the marine realm; however, freshwater species of Saurichthys are also known, particularly from South Africa, Eastern Europe and Australia (e.g. Turner 1982; Kogan et al. 2009; Romano et al. 2012). Whilst the main focus of this work was to explore the taxonomic utility and document shape trends of the opercle on a large sample of specimens referred to species, a worthwhile expansion of these results, which will form the basis of a future contribution, would be to assess whether the axes of opercle shape change so far identified are also present among freshwater species, and whether changes in shape documented between the two realms in extant species can also be identified for Saurichthys.

Over the course of development from young larva to full adult stage in sticklebacks, opercle shape also changes along the same axis indicating that a paedomorphic truncation of development underlies the evolution of opercle shape in sticklebacks (Kimmel et al. 2012b). Uniting morphometric measurements of opercles for juvenile specimens of Saurichthys with those data collected here on adult specimens would provide an opportunity to evaluate the evolution of development in the fossil record (e.g. Raff 2007; Sánchez-Villagra 2012; Wilson 2012b), thereby considering the temporal persistence of the developmental processes already proposed to determine opercular shape evolution (Kimmel et al. 2012b). Few well-preserved opercles from identified juvenile specimens are available for comparison; nevertheless, Rieppel (1992) already noted the shape of the operculum in two juveniles of S. curionii (PIMUZ T 4451 and T 4452) appeared somewhat more oval in shape, compared to the adult more rounded shape, suggesting ontogenetic shape change may occur most conspicuously along an anteroventral axis. We observe very subtle shape difference in the opercle of juvenile (PIMUZ T 309, and T 603) and adult specimens of S. costasquamosus. The less conspicuous difference for S. costasquamosus is most likely due to the later ontogenetic stage of these specimens compared with those of S. curionii examined by Rieppel (1992), in which the opercles were typically one third the length of adult size, compared with one-half for the S. costasquamosus specimens.


In this study, we provide the first quantitative assessment in any Mesozoic fish of the taxonomic utility of linear measurements and geometric morphometric quantification of opercle size and shape. Our results indicate species-specific shape change and subtle intraspecific variability in the opercle. We show that interspecific shape variation is largely explained by a dimension decrease on a horizontal plane of the anterior and posterior margins of the opercle coupled with perpendicular increase in dimension of the dorsal and ventral edges, whereas more subtle differences in shape can be detected along all margins of the opercle at an intraspecific level. Our results are sample dependent and should be further tested on a wider proportion of Saurichthys diversity; nevertheless, we underline the broad applicability of these methods to other groups. We document modes of shape change in the opercle and discuss how insights from morphofunctional and ontogenetic studies of extant fish may help interpretation of opercle characters in a future, necessary taxonomic revision of the genus Saurichthys.


We thank Carlo Romano (Zürich), Ilja Kogan (Freiberg) and Walter Salzburger (Basel) for insightful discussion and comments. An anonymous reviewer and Adriana López-Arbarello are thanked for valuable comments on a previous version of this paper. Thomas Brühwiler, Julia Huber, Heinz Lanz, Urs Oberli, Christian Obrist, Leonie Pauli and Sergio Rampinelli are thanked for preparation of fossils. We are grateful to Madeleine Geiger and Rosi Roth (both Zürich) for help with photography. Monette Véran (Natural History Museum, Paris), Ursula Göhlich (Natural History Museum, Vienna) and Irene Zorn (Geological Survey, Vienna) are thanked for their help with access to collections. This work was supported by the Swiss National Science Foundation (SNF) Sinergia programme to Marcelo R. Sánchez-Villagra, Heinz Furrer and Walter Salzburger (CRSII3-136293).