Palaeoecological analysis

It is essential for the interpretation of morphological variability in fossil bivalve shells to know as much as possible about their habitat. This includes data on co-occurring organisms, community structure, and trophic webs. The macrofossil collections from the sampled horizons are qualitative in nature and do not depict relative abundance of the taxa. Nevertheless, wherever possible, the fossils from each particular horizon have been assigned to a community or association each of which is characterised by specific ecological requirements (Table 1). Most of the fossil assemblages could be assigned to distinct faunal associations as defined by Fürsich (1981a) and Fürsich and Werner (1984, 1986, 1991). For a few samples new associations had to be defined. Additionally, the microfauna and/or -facies of most of the sampled horizons were investigated briefly. From argillaceous and muddy layers sediment samples were processed by immersion in diluted hydrogen peroxide, wet-sieved down to a mesh size of 0.01 mm, dried, and picked for microfossils. From well-cemented rocks thin-sections were prepared and analysed using a binocular microscope.

Stable oxygen isotopes

Since the relationship between stable oxygen isotopes of carbonate skeletons and ambient seawater was discovered (Epstein et al. 1951), δ¹⁸O analyses on carbonates have become a standard tool for reconstruction of (palaeo-) temperatures (e.g., Sharp 2007). In the present study, oxygen isotope analyses were carried out to gain additional information on the Late Jurassic palaeoclimate in Portugal; these analyses help to identify the factors that possibly influenced bivalve shellshape. Analysis of δ¹⁸O was performed on the primarily calcitic shells of the oysters Praeexogyra, Actinostreon, and Nanogyra from different strata across the Lusitanian Basin, and on three belemnite rostra (Hibolites sp.) from the Abadia Formation. All shells were previously screened for diagenesis, applying cathodoluminescence analysis and atomic absorption spectrometry. Only non-luminescent portions of clean, shiny, and translucent calcite containing < 250 ppm Mn and < 700 ppm Fe were chosen for analysis. For details on sample localities, preparation technique, and control for diagenesis see Schneider et al. (2009). Samples for stable isotope analysis were milled using an agate mortar and pestle, weighed, and measured at the laboratory of the GeoBioCenter at the Palaeontology Section of the Ludwig-Maximihans-University Munich, using a Delta plus Thermo/Finnigan MAT mass spectrometer coupled with a GasBench II preparation device. Measured values are adjusted to solid reference standards NBS-18, NBS-19, and “Pfeil” internal lab standard. Accuracy generally ranges between 0.1‰ and 0.2‰. Palaeotemperatures were reconstructed using the equation of Anderson and Arthur (1983): T [°C] = 16.0 - 4.14(δ¹⁸O_carbonate - δ¹⁸O_water) + 0.13 (δ¹⁸O_carbon - δ¹⁸O_water)² assuming ice free poles (δ¹⁸O_water = -1.2‰ PDB) and normal marine salinity (36 psu). The latter was evaluated by the presence or absence of stenohaline organisms (e.g., brachiopods, corals, echinoderms).

Sample preparation and measurements

After chemical and/or mechanical preparation, the length and height of all individuals of the three target taxa were measured using a calliper (measuring accuracy: 0.5 mm; Fig. 6). For Isognomon, the length of the ligament area and number of resilifers were recorded. Most of the specimens are incomplete and large portions of the ventral part of the shell are missing. Length, height, and thickness were recorded at the latest entirely traceable growth line. As the completeness of the valves usually varies within a single double-valved specimen, the more complete valve of each specimen was chosen for analysis. Because the umbo is commonly broken in Arcomytilus and Isognomon, measurements for length and ligament length were reconstructed for missing parts of up to 10 mm in length. Altogether, 364 valves of Arcomytilus, 638 valves of Isognomon, and 385 valves of Eomiodon were measured. Supplementary to the measurements, the number of ribs of Arcomytilus was counted in 289 valves. Counting was done along a clearly traceable growth line at about the stage of maximum shell growth (?maturity; arrow in Fig. 6A). Although several phases of slightly reduced shell growth may occur during the life of a mussel, the turning point, when growth starts to decrease substantially, can be determined quite accurately in a well-preserved, adult individual, and may represent the best point to compare rib numbers in specimens of a similar ontogenetic stage.

Fig. 6.

Measured distances in the three target taxa. A. Arcomytilus. B. Isognomvn. C. Eomiodon. Abbreviations: H, height; L, length; LL, ligament length. Arrow indicates turning point of growth.

Outline analysis

For outline analysis, specimens were oriented in such a way that the plane of the latest complete growth stage was horizontal and then photographed. Digital images were reduced to this growth stage using the rubber tool in Adobe® Photoshop CS. Subsequently, the contrast of the images was increased to obtain black and white images. The contours of the shell outlines were saved as x/y-coordinates using TpsDig 2.12 software (Rohlf 2008). In Arcomytilus and Isognomon, the tracing of x/y-coordinates was started at the umbo. In Eomiodon, the centre of the lunule, which is usually more sharply defined as the umbo, was chosen as the starting point. As normalisation for size (executed by the Hangle software; see below) is included in the procedure of FFT (Fast Fourier Transformation; see details below), all coordinates registered by the software were kept for outline analysis. In total, data sets of 364 Arcomytilus, 561 Isognomon, and 321 Eomiodon specimens were produced.

Fourier shape analysis is a well-established method for analysing outline shape based on a combination of sine and cosine waves (Kuhl and Giardina 1982; Crampton and Haines 1996) that has been successfully tested on various biologic objects (e.g., Renaud and Michaux 2004; Bond and Beamer 2006; Liow 2006; Ponton 2006; Schulz-Mirbach and Reichenbacher 2008) including several Cretaceous and Cainozoic bivalve groups, some of which display shell shapes similar to the target taxa (Crampton and Maxwell 2000; Haines and Crampton 2000; Palmer et al. 2004; Scholz 2003; Crampton and Gale 2005, 2009; Scholz and Hartman 2007a, b). Different methods for this procedure have been developed, from which FFT was chosen because of certain advantages discussed at length by Haines and Crampton (2000). The data sets were processed using the programs Hangle, Hmatch, and Heurve designed by Crampton and Haines (1996). The outlines were automatically smoothed (12 times) in order to eliminate pixel “noise” and insignificant outline details resulting from minor shell breakage. The numbers of harmonics that sufficiently describe the outlines were evaluated from amplitude vs. harmonic number plots according to Lestrel (1989). These plots indicated that at least seven Fourier harmonics were necessary for the analysis of Arcomytilus, eight for Isognomon, and nine for Eomiodon. Subsequently, the data sets were normalised for starting positions using the program Hmatch, because the outlines of all three taxa lack sharply defined homologous points. By inversion of the Fourier transformation, idealised synthetic shell outlines were generated for the first two principal components using the program Heurve and following the approach outlined by Haines and Crampton (2000).

Statistical analyses

Height/length ratios, number of ribs, and size distribution patterns were analysed using SPSS 16.0 (SPSS Inc. 2007). All taxa were tested for correlation between the measured size parameters. In a first step, scatter plots of log transformed values of height vs. length were produced. Additionally, size was displayed in box plot diagrams where each box corresponds to one lithostratigraphic unit. The measurements are generally based on adult specimens, which can be identified by the crowding of growth lines in mature individuals (see Figs. 3–5 for examples), but also include a minor portion of juvenile or sub-adult individuals, which account for negative outliers and may enlarge the extension of the lower ranges of the boxes.

The Fourier coefficients were analysed with regard to stratigraphic, palaeoecologic, and/or taxonomic aspects applying Principal Component Analysis (PCA) based on the variance-covariance (VCV) matrix using SPSS 16.0 and PAST 1.82b software packages (Hammer et al. 2001). In order to avoid confusion caused by too many symbols, convex hulls were plotted for all samples with N ≥ 3. Smaller samples were only included in the plots in case they enlarge the total morphospace, and thus play a role for the interpretation of the diagrams. The number of relevant principal components, i.e., those accounting for more variance than expected to be explained by random, was evaluated applying the broken stick model sensu Jackson (1993). Generally, variability was found to be best displayed in all cases by plots of PC2 vs. PC1. In neither of the analyses, PC3 helped to obtain a better decomposition of groups when plotted vs. PC1 or PC2, nor to better explain morphologic differences; thus we refrain from discussing and/or figuring these plots.

Potential evolutionary changes may be best inferred by grouping the data according to lithostratigraphy. For this purpose, plots of PC1 vs. PC2 have been constructed and groups displayed as convex hulls. Additionally, a series of artificial shell outlines was plotted on these diagrams to illustrate the morphospace of the taxa. Moreover, the confidence ellipses of group means and corresponding mean shell outlines have been produced and figured separately. In all three taxa, the lithostratigraphic groups are by far best separated by plotting the first two principal components.

In addition to lithostratigraphic grouping, Fourier coefficients were displayed in separate population plots, each representing a single time-slice or lithostratigraphic unit, in order to detect ecophenotypic adaptations and evaluate their potential relationship to biogenic and non-biogenic factors (see Supplementary Online Material [see  http://app.pan.pl/SOM/app55-Schneider_etal_SOM.pdf for SOM: figs.S1–S10]). The term “population” is used herein for all specimens of a taxon derived from a single horizon. Most of these samples may not represent populations in a biological sense, but were subject to more or less distinct time-averaging (Walker and Bambach 1971; Fürsich and Aberhan 1990; Kidwell and Bosence 1991; see Fürsich et al. 2009 for details). The abbreviations used for the populations are listed in Tables 1 and 2 (column 1).

Assemblages and associations

Analysis of the samples reveals a number of macrobenthic assemblages falling fully within or deviating slightly from the associations already recognised by Fürsich (1981a) and Fürsich and Werner (1984, 1986, 1991) (Table 1). The term Isognomon rugosus association is used also for associations containing Isognomon lusitanicus herein (see Systematic palaeontology section below) in order to provide palaeoecological comparability. This association is represented by several samples, attributable to at least five subsets, one of which, termed Isognomon lusitanicus/Cylindrobullina sp. subset, is newly described herein. This subset, recorded from the Sobral member at A dos Arcos [4A] and Chão da Cruz [4C], is composed almost exclusively of I. lusitanicus attaining a large size; the specimens are sparsely encrusted with Praeexogyra pustulosa; solitary corals (cf. Axosmilia sp.) may occur accessorily. The subset occurs in marly, finegrained bioclastic sediment that contains sparse microfauna of small planispiral lituolid foraminifera, few miliolid foraminifera and several ostracod species (e.g., Cytherelloidea sp., Paracypris sp.). This assemblage together with the occurrence of two taxa of small gastropods, i.e., cf. Cylindrobullina sp. and a neritimorph, indicates euhaline or slightly brachyhaline salinities. In addition, Isognomon rugosus and I. lusitanicus were sampled from several horizons containing parautochthonous accumulations of disarticulated shells (termed “parautochthonous I. rugosus assemblage” herein).

The Eomiodon securiformis/nerineid sp. B association, found at Chão da Cruz [4B, D], Arranhó [4E, F], Nossa Senhora do Ajuda [4G] (all from the Sobral member [4], Farta Pao formation, Arruda Subbasin), and at Sobral da Lagoa [3J] (Alcobaça formation [3], Bombarral Subbasin), is newly defined herein. Typically, this association is dominated by E. securiformis and a relatively small, moderately high-coiled nerineid gastropod (nerineid sp. B of Fürsich and Werner 1986). These are frequently accompanied by large Amauropsis sp., Fuersichella bicornis, and Anisocardia (Antiquicyprina) sp. B (sensu Fürsich and Werner 1986). Additionally, a variety of different bivalve taxa and several corals may occur.

Several new assemblages are recorded only from single outcrops and therefore cannot be classified as associations sensu Fürsich (1981a). The autochthonous Arcomytilus morrisii/Corbulomima suprajurensis assemblage occurs in dark, marly clay at Lameiro das Antas [6M] (Arranhó II member [6] of the Farta Pao formation, Arruda Subbasin). It is characterised by abundant A. morrisii occurring in dense clusters, by the presence of a few specimens of the large isognomonoid Rostroperna thurmanni, a relatively low-diversity fauna of small endobenthic bivalves dominated by Corbulomima suprajurensis, and a remarkably diverse echinoderm fauna (isolated skeletal elements) including common ophiuroideans.

The limestones at the Cesareda Plateau (Cabaços formation [1], Bombarral Subbasin) yielded two distinctive faunal assemblages that previously had not been encountered. The autochthonous Regulifer beirensis assemblage [1A] is dominated by crowded specimens of the large, thick-shelled mytilid Regulifer beirensis, associated in a few cases with Isognomon rugosus and the epibyssate pectinid Camptonectes auritus. Further up-section, greyish wackestones contain a parautochthonous chaetetid-terebratulid-Arcomrytilus assemblage [1B]. Small (< 5 cm) irregularly globular chaetetid colonies, terebratulid brachiopods, and disarticulated Arcomytilus shells dominate the likely parautochthonous assemblage.

A few Arcomytilus specimens occur in a fully marine endo-/epibenthic assemblage of rather exceptional composition for the Lusitanian Basin, which is found south of Carrascal [3O] (Alcobaça formation [3], Bombarral Subbasin). The fauna is predominantly composed of large burrowing anomalodesmatan (Pholadomya, Ceratomya) and heterodont bivalves, accompanied by the epibenthic reclining astartid Coelastarte discus (see Schneider and Werner 2007).

Several samples (termed “isolated specimens” in Table 1) consist of single or just a few specimens found isolated or within assemblages that could not be sensibly addressed. These samples often stem from less well-documented regions or time slices, and are included in the analyses to fill these gaps. Similarly, the samples from the IGM cannot be assigned to distinct faunal assemblages/associations with confidence because details on sampling procedure and species composition are lacking.

Stable oxygen isotopes

The measured δ¹⁸O values range from -2.84 to 1.18 VPDB. The palaeotemperatures reconstructed from these values range from 7 to 23°C. However, in several samples that derive from brackish water environments, freshwater mixing may have significantly influenced the δ¹⁸O values resulting in a deviation of several degrees from original temperature (Sharp 2007). Additionally, several samples that produced significant results for Sr-isotope stratigraphy (Schneider et al. 2009) did not yield equally significant oxygen values, as revealed by comparison of two or more samples from the same horizon. Consequently, only those values are considered for temperature reconstruction that (i) were obtained from faunal assemblages indicating fully marine conditions and (ii) are confirmed by two or more samples from the same horizon yielding similar isotope values. The palaeotemperatures reconstructed from the oyster samples suggest 16–22.5°C for the Late Oxfordian—Early Kimmeridgian (14 samples) and 15.5– 19°C for the Late Kimmeridgian (5 samples). Three belemnite samples from the Lower Kimmeridgian Abadia Formation yield isotope values indicating 12–14°C.

Fig. 7.

Scatter plots of log transformed values of height over length for the three target taxa. A. Arcomytilus. B. Isognomon. C. Eomiodon. Numbers in squared brackets refer to Fig. 2.

Morphometry

Size and size increase.—The height/length plots of all taxa show a distinct linear correlation (Fig. 7), which, however, is far more constant in Eomiodon than in Arcomytilus and Isognomon. The most obvious evolutionary process that occurs in all three genera studied is a sudden size increase (Fürsich and Werner 1986) that took place shortly after the Early/Late Kimmeridgian boundary, relatively precisely timed at ∼152 Ma based on Sr isotope ages (Schneider et al. 2009). However, a distinct separation of any group is impossible based on the height/length diagrams, although the Isognomon specimens that are preserved as moulds (Cabo Mondego formation [2], Cabaços Formation [1], “São Martinho” [3A–D]) exhibit a slight shift to higher y-values compared to specimens of similar size that are preserved with shell (Fig. 7B). Therefore, size increase is illustrated by box plot diagrams for all genera (Fig. 8). One-way ANOVA reveals significance for Arcomytilus (F_{[8; 353]} = 69.2; p < 0.001), Isognomon (F_{[4; 295]} = 162.8; p < 0.001), and Eomiodon (F_{[4; 380]} = 90.1; p < 0.001) at the 0.05 level.

For specific reasons, the box-plots are based on different size parameters. All species of Arcomytilus in general and Arcomytilus morrisii in particular vary considerably in shape; often, long specimens are characterised by a relatively small height and vice versa. In order to obtain a best estimate for size, the box-plot for Arcomytilus is based on the log transformed geometric mean of length and height. Three boxes that display additional data sets of closely related species of Arcomytilus from the Jurassic of France are included in this plot. Based on a Bonferroni post hoc test, the French Bathonian species Arcomytilus bathonicus and Arcomytilus asper are distinctly smaller than A. morrisii from Portugal (Figs. 3, 8A). The data set of A. pectinatus contains scattered data and is included only to illustrate potential ancestry to A. morrisii, as discussed by Fürsich and Werner (1988). Within the Upper Jurassic data sets of A. morrisii, the samples corresponding to the Cabo Mondego [2], Cabaços [1], and Alcobaça formations [3] and the Sobral [4], Arranhó I [5], and Arranhó II [6] members are clearly separated based on the Bonferroni test (p < 0.001), although their whiskers show a considerable overlap of absolute sizes due to the inclusion of sub-adult specimens.

The box-plot of Isognomon is based on log transformed ligament length, because the thin ventral part of the shell is commonly broken off in specimens from Portugal, and corresponding measurements of shell length and height are therefore often taken from sub-adult growth stages. With exception of a few gerontic specimens that usually exhibit strongly allometric growth favouring height, ligament length is directly correlated with shell length. The plot shows a distinct gap between the boxes of the Cabo Mondego [2] and Alcobaça formations [3] (no specimens with shell preservation could be obtained from the Cabaços Formation [1]) and those of the different members of the Farta Pao formation [4, 5, 6] (Fig. 8B), which is confirmed to be significant by a Bonferroni post hoc test (p < 0.001). Again, the whiskers indicate a considerable overlap of absolute sizes, which may, however, in this particular case be due largely to juvenile specimens (especially for the Arranhó I [5] and Arranhó II [6] members). The few positive outliers from the Alcobaça formation [3] that prevent the total separation of the two groups based on size of adult specimens were collected from a single horizon at Consolação [3F]. With regard to lithostratigraphy, this horizon is attributed to the Upper Oxfordian—Upper Kimmeridgian Alcobaça formation. However, sample [3F] is positioned near the top of this formation and Sr-isotope data indicate a Late Kimmeridgian age, which appears to postdate the onset of size increase. Based on these considerations, Isognomon from the Upper Jurassic strata of the Lusitanian Basin may be separated into two chronostratigraphically distinct size groups, Oxfordian—Lower Kimmeridgian and Upper Kimmeridgian—Tithonian, based on ligament length. Remarkably, the same horizon [3F] yielded a considerable number of Arcomytilus shells that clearly belong to the group of minor size. Certainly, size increase occurred within a distinct time span but not absolutely simultaneous in different bivalve taxa.

Fig. 8.

Box plots of size for the three target taxa. A. Size of Arcomytilus based on log transformed geometric means of length and height. B. Size of Isognomon based on log transformed ligament length. C. Size of Eomiodon based on log transformed shell length. Arrangement of boxes corresponding more or less to their stratigraphic succession, from left to right. Numbers in squared brackets refer to Fig. 2.

Fig. 9.

Scatter plot of resilifer number over ligament length in Isognomon. The two groups that correspond to lithostratigraphy are clearly visible. Numbers in squared brackets refer to Fig. 2.

In Eomiodon, size increase is also evident, but occurs less abruptly than in Arcomytilus and Isognomon. The samples stem from four units only (Cabo Mondego [2] and Alcobaça [3] formations and Sobral [4] and Arranhó II [6] members). The exclusively carbonate rocks of the Cabaços Formation [1] and the Arranhó I member [5] formed under conditions most likely unsuitable for Eomiodon. A sample of the Choffat collection at the IGM from Cabo Espichel is treated separately in this analysis, as it cannot be assigned to one of the Farta Pao members with confidence. Although Eomiodon may casually be subject to moderately allometric changes during ontogeny, these modifications are not clearly displayed in the height/length ratio; consequently, size is relatively accurately displayed by length, and the plot is based on log transformed values of this parameter (Fig. 8C). The Bonferroni post hoc test confirms a significant separation of the Cabo Mondego [2] and Alcobaça [3] samples from those of the Farta Pao formation (p ≤ 0.002). Apart from the taxa studied herein, a more or less contemporary intraspecific or intrageneric size increase can be observed in several other bivalve taxa from the Lusitanian Basin (e.g., Jurassicorbula spp., Coelastarte discus, Myophorella spp.; FTF, SS, WW personal observations), indicating that this process is fairly widespread.

Resilifer density in Isognomon. —In Isognomon, the step in size increase co-occurs with a substantial change in resilifer density that is displayed in an additional scatter plot of the number of resilifers against ligament length (Fig. 9). Naturally, this plot is only based on specimens with shell-preservation. Although a few outliers exist, two groups that correspond to the Late Oxfordian—Early Kimmeridgian (Cabo Mondego [2] and Alcobaça [3] formations; Group I) and the Late Kimmeridgian—Early Tithonian (Farta Pao formation [4, 5, 6]; Group II) can be relatively clearly separated based on the plot. Sample [3F] is represented by a single specimen in this plot, as all other specimens from this sample are articulated, and the number of resilifers cannot be counted; this single specimen, however, plots within Group I (as do four other specimens from the Farta Pao formation). The change in resilifer density occurs as follows: While total size and respective ligament length rise, the number of resilifers stays constant, and thus the number of resilifers per cm decreases. Based on these observations, a separation of Isognomon from the Late Jurassic of Portugal into two chronologically different species seems evident. Including data from the figured type specimens of Isognomon rugosus (Münster, 1835) and Isognomon lusitanicus (Sharpe, 1850) these specimens clearly resemble the two groups of resilifer density. Consequently, Group I may be correctly identified as I. rugosus (Münster, 1835), while for Group II the original identification as I. lusitanicus (Sharpe, 1850), considered by Fürsich and Werner (1989a) as a synonym of I. rugosus, is re-established (for details on the taxonomy, see Systematic palaeontology section below). Several specimens from the Upper Jurassic of France, deposited at the MNHN and labelled Isognomon rugosus and Isognomon bergeroni, respectively, are positioned next to the right margin of the rugosus-group, but still are included therein.

Rib numbers in Arcomytilus.—In Arcomytilus, the total number of radial ribs represents an additional parameter that may be utilised to infer evolutionary changes. Whereas Arcomytilus bathonicus and Arcomytilus asper, co-occurring in Upper Bathonian sediments at Luc-sur-Mer (Calvados, France), can be separated clearly based on rib numbers (Fig. 10), Arcomytilus asper, Arcomytilus pectinatus, and Arcomytilus morrisii from Portugal cannot be distinguished from each other by this method. Interestingly, the specimens from the Cabo Mondego formation [2] are significantly different from those from the Arranhó I [5] and Arranhó II [6] members based on one-way ANOVA (F_{[8; 286]} = 23.1; p < 0.001) and a Bonferroni post hoc test (p < 0.001), while they show similar rib numbers (p = 1.0) to those from the Sobral member [4] (Fig. 10). All in all, a largely continuous, gradual decrease of rib numbers from the Upper Bathonian A. asper to the Oxfordian to Kimmeridgian A. pectinatus and across the whole range of A. morrisii is inferred and a clear separation of two or more groups of Arcomytilus from the Upper Jurassic of Portugal based on rib numbers is impossible. Moreover, an increase in rib number in specimens from the Upper Kimmeridgian Sobral member [4] is against the trend.

Fig. 10.

Box plot of rib numbers in Arcomytilus. Numbers in squared brackets refer to Fig. 2.

Fig. 11.

Lithostratigraphy plot of Arcomytilus. A. Different species and lithostratigraphically grouped Arcomytilus morrisii are displayed as convex hulls. Calculated artificial shell outlines for full number coordinate pairs are plotted to illustrate the morphospace. M; mean artificial shell outline. B. 95% confidence ellipses of group means and corresponding calculated shell outlines for group means are plotted. Numbers in squared brackets refer to Fig. 2.

Table 3.

Number of shell outlines and percentage of explained variance of principal components 1–3 for PCA plots in Figs. 9–11 and S1–S10.

Fig. 12.

PCA plot of shell shape in Arcomytilus, grouped according to rib number in steps of 25 ribs and displayed as convex hulls.

Shell outline shape.—Height/length plots for the three bivalve genera clearly show that apparent morphological variability of the shells cannot be sufficiently displayed by such a simple method (Fig. 7). The same holds true for any other combination of linear measurements tested, including inflation, ligament length or height, or absolute shell thickness. In contrast, the PCA plots of Fourier coefficients display the major part of information on shell outline shape retained in the analyses. In all analyses, not more than the first three principal components are relevant based on broken stick analysis (Jackson 1993), and the first one (PC1) accounts for > 29% of total variance. Cumulative variance of PC1–3 generally ranges between 59 and 70%. The percentages of explained total variance for PC1, PC2, and PC3 in all PCAs are listed in Table 3.

In Arcomytilus, the total range of shape variability of A. morrisii expands notably with time and size. Slender, sub-elliptical or tapering forms that dominate the stocks of the older strata and also the three species from the Middle and Upper Jurassic of France are joined by broadened or high sub-triangular shells in the Arranhó I [5] and Arranhó II [6] members (Fig. 11A). All convex hulls overlap considerably, rendering a sensible separation of any group impossible. As can clearly be inferred from the plot, shape and size are not correlated, since the morphospace occupied by the large specimens from the Sobral member [4] is largely included in those spanned by the small Cabo Mondego [2] and Alcobaça [3] specimens. Almost the entire morphospace of Arcomytilus morrisii is covered by the specimens from the Arranhó I member [5]. The ellipse plot of group means shows variability at a fairly small scale (Fig. 11B). A truly consecutive tendency of shape change is interrupted by the specimens from the Alcobaça formation [3], which plot between the group means of the stratigraphically older Cabaços [1] and Cabo Mondego [2] formations. It should be noted that the low number of seven harmonics readily describes the general differences in shell shape, but does not capture shape details of the umbo, which are quite different among species and lithostratigraphic groups. Generally, Arcomytilus morrisii is characterised by tapering, pointed, slender umbones, while these are much blunter in Arcomytilus bathonicus, Arcomytilus asper, and Arcomytilus pectinatus (Fig. 3). Additionally, specimens of A. morrisii from the Cabaços [1] and Cabo Mondego [2] formations typically have slightly more blunt umbones than the individuals from younger strata.

Fig. 13.

Lithostratigraphy plot of Isognomon. A. Lithostratigraphically arranged groups are displayed as convex hulls. Calculated artificial shell outlines for full number coordinate pairs are plotted to illustrate the morphospace. M, mean artificial shell outline. B. 95% confidence ellipses of group means and corresponding calculated shell outlines for group means are plotted. Numbers in squared brackets refer to Fig. 2.

Fig. 14.

Lithostratigraphy plot of Eomiodon securiformis. A. Lithostratigraphically arranged groups are displayed as convex hulls. Calculated artificial shell outlines for full number coordinate pairs are plotted to illustrate the morphospace. M, mean artificial shell outline. B. 95% confidence ellipses of group means and corresponding calculated shell outlines for group means are plotted. Numbers in squared brackets refer to Fig. 2.

In order to evaluate the relationship between rib number and shell shape, the Fourier coefficients are displayed in a separate plot grouped by rib number, with each group comprising a 25 ribs interval. The plot shows large to complete overlap and almost identical centres of all five convex hulls, and a separation of any group is not indicated (Fig. 12).

In Isognomon, a separate group of shell outlines, termed “São Martinho”, has been included in the lithostratigraphy PCA plot (Fig. 13). The data gathered in this group comprise two large samples from the coastal outcrops at São Martinho do Porto and Salir do Porto that form a single horizon exposed at either end of the cove. Based on lithology and Sr isotope data, these outcrops are assigned to the Alcobaça formation (Schneider et al. 2009). However, as nearly all specimens from the Cabo Mondego and Cabaços formations, these Isognomon are preserved as partial moulds. Since the aragonitic inner part is much thickened at the anterior side of an Isognomon shell due to shell weighting (compare Seilacher 1984) the thin outer calcitic layer has been compressed and imprinted onto the internal mould when the aragonite was dissolved during diagenesis. As a result, the shell outlines are artificially shortened (as can also be seen from the height/length plot) and cannot be directly compared to specimens in full shell preservation. However, the shell outlines of the partial moulds can still be compared among different samples with similar preservation. The convex hulls of specimens from the Cabo Mondego [2] and Cabaços [1] formations and from São Martinho [3A–D] display almost complete overlap (Fig. 13A). Despite the shortening artefact that results in a shift towards higher PC 1 values, all three groups show more than 50% overlap with the convex hulls of groups with shell-preservation, suggesting that originally there was no major difference in the occupied morphospace. Additionally, the convex hulls of groups with preserved shell overlap to a large degree (Fig 13A) while group means, except that of Arranhó II [6], are almost equal (Fig. 13B) and thus a separation of the previously defined Groups I and II, indicated by ligament length and resilifers per cm, cannot be deduced from shell outline shape. Moderately increased shape variability including specimens with a more triangular shell outline (indicated by declining PC1 values), more tapering beaks (decreasing PC2 values), and higher ligament area (increasing PC2 values) is displayed by the younger and larger shells from the Farta Pao formation.

Eomiodon displays somewhat less variability than the other genera, not only with regard to size but also shape (Fig. 14A). Moreover, separation of different lithostratigraphic groups based on shell outline shape is impossible. The convex hulls of specimens from the Alcobaça [3] and Sobral [4] formations overlap to a great extent. The Arranhó II [6] group occupies a relatively small area in the centre of the Sobral morphospace [4] and is nearly entirely included in the Alcobaça group [3]. In contrast to Arcomytilus and Isognomon, where shape variability increases with time, the largest variability in Eomiodon is observed in the stratigraphically oldest specimens from the Cabo Mondego formation [2]. The latter group includes several more elongated specimens (decreasing PC2 values) either with well-projecting, strongly prosogyrous (decreasing PC1 values) or slightly elevated, almost central beaks and a narrow, slightly tapering posterior shell (increasing PC1 values). Its convex hull overlaps only for about 50% with the others. In contrast, specimens from the Alcobaça formation [3] and Sobral [4] and Arranhó II [6] members have rather uniform, relatively high and short shell outlines with markedly prosogyrous umbones, relatively straight ventral margin, and a generally more equal-sided triangular shape. The confidence ellipses of the latter two group means are almost concentric. The ellipses of the Alcobaça and Cabo Mondego formations are centred slightly below (Fig. 14B).

Three population plots for different lithostratigraphic units or time slices have been produced from the Fourier coefficients of Arcomytilus (SOM: figs. S1–S3), four of Isognomon (SOM: figs. S4–S7), and another three of Eomiodon (SOM: figs. S8–S10). In all of these plots, individual populations show large areas of overlap, and the convex hulls of large samples generally span a major part of the total morphospace. In a few cases, particular samples exceed the total morphospace in a certain direction. For example, Arcomytilus populations [4R] and [4T] in SOM figure S2 include straight, elongated specimens that extend the morphospace towards higher PC1 and PC2 values. In SOM figure S8, the morphospace of the Salgados Eomiodon population [3M] enlarges towards lower PC1 and PC2 values due to a single, strongly elongated, gerontic individual. However, no population is clearly separated from all others in any of the plots.

Samples assignable to the same faunal associations may occupy a similar morphospace, e.g., Arcomytilus populations [4D, 4F] that belong to the Eomiodon securiformis/nerineid sp. B association (SOM: fig. S1), and Eomiodon samples [6E, 6F] from merely identical environments that are assigned to an almost monospecific Eomiodon securiformis association (SOM: fig. S10). Conversely, other populations that belong to identical associations share only part of the morphospace. Although both Isognomon samples are assigned to the Isognomon lusitanicus/Cylindrobullina sp. subset and come from similar facies in the same area of the Arruda Subbasin, population [4C] in SOM: fig. S6 includes a number of more slender, elongated shells, while the stout, sub-rectangular forms included in sample [4A] are absent and vice versa. Moreover, samples assigned to different associations may occupy congruent morphospace parts. While [4B] and [4F] in SOM: fig. S9 represent Eomiodon populations from the Eomiodon securiformis/nerineid sp. B association, and population [4H] stems from an almost monospecific E. securiformis association, the respective convex hulls overlap to a large degree.

Although all convex hulls show a large area of overlap, Isognomon rugosus samples from the Arruda Subbasin in SOM: fig. S7 are dominated by stout shells with high and rounded ligamental areas, while populations from the coastal sections also include more slender and more triangular specimens. In both regions, populations with a similar morphospace belong to different associations and subsets and come from different facies, e.g., [5A] from carbonate rocks and [6J] from sandy marl.

Evolutionary implications

Size increase.—Several potential causes may trigger size increase in bivalves. In the authors' opinion, increasing nutrient supply and enhanced primary productivity is the most likely scenario in the case of the Upper Jurassic Lusitanian Basin, as other factors can be neglected or considered at least highly unlikely.

(1) Evolutionary size increase, commonly known as Cope's Rule, has been recorded for various groups of organisms including Jurassic bivalves (Hallam 1975, 1998; Johnson 1994) and is regarded as “a genuine evolutionary phenomenon” (Hallam 1975: 493) in those animals. Usually, increasing body size of bivalves is correlated with a decline in population size that is limited by the available food resources (Hallam 1998). However, all studied taxa from Portugal formed dense populations up to the Tithonian, as can be inferred from in situ preserved bivalve populations. Moreover, the examples discussed by Hallam (1975, 1998) and Johnson (1994) display a slow (several million years) and gradual size increase, whereas in the taxa from Portugal this process occurs relatively rapidly and in a punctuated pattern, and thus is not evolutionary in the sense of Cope's Rule, but is rather controlled by one or more environmental factors. A similar case is reported by Johnson (1999), dealing with size reduction in Bathonian bivalves. As a result, evolutionary size increase may be a common phenomenon in various organisms but obviously has to be tested for each particular case (e.g., Gould and MacFadden 2004).

(2) The major change from a carbonate to a siliciclastic regime in the Lusitanian Basin directly followed an episode of intense rifting (Rasmussen et al. 1998; Alves et al. 2002) and occurred around the Oxfordian—Kimmeridgian transition (Leinfelder and Wilson 1998). It therefore clearly predates the size increase seen in the bivalves. Anyway, size seems to be independent of facies, as both small and large specimens of all three genera are recorded from carbonate and siliciclastic horizons.

(3) All three taxa generally flourished in brackish waters, but also occurred under more or less fully marine conditions. However, the composition of the different associations is largely driven by salinity (Fürsich and Werner 1984, 1986), and the appearance of identical associations during both time intervals in the Lusitanian Basin excludes salinity as a factor delimiting size.

(4) During the Late Jurassic, Portugal was situated close to the northern tropics (Smith et al. 1994; Blakey 2007) and generally characterised by summer-wet tropical climate and minor seasonal temperature fluctuations, as suggested by general circulation models (Sellwood and Valdes 2006). Moreover, no major climatic changes during the Late Jurassic are evident from this area and estimated mean land-surface temperature based on climate models is around 30°C (Sellwood and Valdes 2006), while sea-surface temperatures range between 20–30°C (Valdes and Sellwood 1992; Moore et al. 1992). Reconstructed palaeotemperatures based on oxygen isotope data from the Portuguese Upper Jurassic are somewhat lower than these predictions and range between 12–14°C based on belemnites and 15.5–22.5°C based on oysters, which is close to the temperatures calculated from belemnites of Kimmeridgian and Tithonian age from Mallorca (14–16°C; Price and Sellwood 1994). Generally, both belemnites (e.g., McArthur et al. 2007a, b; Wierzbowski and Joachimski 2007) and oysters (Kirby et al. 1998; Surge et al. 2001, 2003) are thought to precipitate calcite with δ¹⁸O in close equilibrium with ambient sea-water. An offset of 2–5°C for belemnites and oysters from the same horizon may be explained by different life style (nektonic versus benthic) and is in good accordance with data gathered from similar taxa from various other regions and time-slices (e.g., Anderson et al. 1994; Wierzbowski 2002; Wierzbowski and Joachimski 2007). The total ranges of temperature values for the Late Oxfordian—Early Kimmeridgian and Late Kimmeridgian—Early Tithonian do not vary significantly, which argues against temperature affecting size increase.

(5) The local sources and directions of freshwater run-off may have been largely constant during the Kimmeridgian and Tithonian as major rifting had already ceased at the end of the Late Oxfordian (Leinfelder and Wilson 1998; Rasmussen et al. 1998; Alves et al. 2002). Due to fairly constant climatic conditions (see previous paragraph), no prominent changes in weathering should have occurred. Consequently, major changes in sea water chemistry resulting from freshwater run-off most likely can be disregarded. Possibly, freshwater discharge into the subbasins that formed during the Late Jurassic due to the diapiric rise of salt pillows (Wilson et al. 1989; Alves et al. 2002, 2003) may have originated from different sources, causing slight differences in water chemistry. However, both small and large bivalve forms occur in two or more subbasins. Thus, local and spatial variations in water chemistry as documented by a carbonate versus siliciclastic sedimentation regime obviously did not affect size in the bivalves. Extensive sea floor spreading in the Tagus Abyssal Plane during the Late Jurassic (Wilson et al. 1989; Alves et al. 2006) may have influenced sea water chemistry (inorganic carbon; e.g., Katz et al. 2005) in the region. However, the partial isolation of the Lusitanian Basin from the open sea renders this scenario unlikely.

(6) Potentially, bivalve size increase may have been triggered by an elevated and more stable food supply, which means higher phytoplankton productivity (Johnson 1994, 1999 and references therein; Perez Camacho et al. 1995). As most bivalves are filter-feeders, they profit more directly from such an increase than, for example, browsing gastropods or echinoderms, which do not show a substantial size increase in the basin during the same time interval (cf. Vermeij 1990). A global increase of phytoplankton productivity during the Late Jurassic is indicated by increasing δ¹³C ratios (Katz et al. 2005; Martin et al. 2008). Whether the Lusitanian Basin was affected by this tendency in spite of its poor connections to the Proto-Atlantic cannot be proven. We have measured δ¹³C ratios of the oyster shells used for palaeotemperatures reconstruction. However, these values are influenced by too many factors (e.g., vital effects, salinity) that cannot be controlled exactly and thus do not allow sound conclusions.

We suggest that progradation of basin infill and continental environments, respectively, may have resulted in an increased nutrient supply by freshwater run-off from these areas into the diminishing marginal marine environments within the Lusitanian Basin. This hypothesis is corroborated by a considerable increase in the diversity and abundance of Dasycladales, Foraminifera, and Ostracoda during the Late Kimmeridgian and Tithonian (Ramalho 1971).

Morphological changes.—Besides size, shell shape is the second important parameter that may yield even more information on the life of a bivalve. As reported above, all target taxa exhibit considerable variability in shell outline. Trying to explain these differences, the life style of the bivalves has to be considered. Arcomytilus, like most Mytilidae, exhibits an anteroventral gape; although this is not a true byssal gape (Mikkelsen and Bieler 2008), it may be inferred from this structure that the bivalves were fixed to the substrate by byssal threads (Cox et al. 1969). Usually, mytilids fix to hard substrates (rocks, hardgrounds, shells) or (bio-)clasts on sand or gravel bottoms (e.g., Tebble 1966; Stanley 1972; Poppe and Goto 1993). Apart from minor morphological modifications for adaptation to an irregularly shaped surface, these individuals are commonly characterised by a straight anterio-ventral shell margin (Stanley 1972), as seen in most specimens of Arcomytilus asper, Arcomytilus bathonicus, and Arcomytilus pectinatus, and also in many Arcomytilus morrisii. The latter species, however, acquired a partially semi-infaunal, endo-byssate life-style, settling on soupy sediment (Fürsich 1980), similar to Quaternary Brachidontes from Argentina (Aguirre et al. 2006) and many other extant and fossil mytiliform bivalves (Stanley 1972). The tapering beaks of these forms were buried within the sediment, while the anterio-ventral portion of the shell formed a large flat surface resting on the sea bottom. Both adaptations helped to stabilise the bivalve in growth position.

A triangular shape, relatively thin shell (the latter cannot be proven for specimens from the Jurassic of Portugal, because of limited preservation quality), and possibly also a larger size may be advantageous in quiet lagoonal settings, as observed for extant Mytilus (Akester and Martel 2000). Moreover, an individual may have a better access to the water column if it develops a more triangular shell, and thus may have an advantage when competing for food. In favourable environments, Arcomytilus morrisii formed dense populations, usually in the form of large clusters composed of numerous individuals, similar to modern Mytilus edulis (which is, of course, notably smaller and strictly epibyssate) on tidal flats of the modern North Sea (SS personal observation).

An additional cause for shape variability in Arcomytilus may have been competition for space (Leinfelder 1986; Aguirre et al. 2006). Usually, the bivalve clusters are composed of individuals of the same generation and therefore of almost equal size. As all individuals grow at about the same rate, space may have become limited within the cluster, causing irregularly shaped shells that enforce growth towards non-occupied patches of the seafloor. However, mytilids generally can cast off and replace byssal threads allowing for limited mobility (Cox et al. 1969). This may explain why individuals of even extreme shapes have normal proportions. Moreover, influence of competition for space on shape variability in modern Mytilus is low (Akester and Martel 2000), and thus may have been also of minor importance in Arcomytilus shells.

The reasons for shape variability in Isognomon may be similar to those in Arcomytilus. The genus Isognomon, defined by an edentulous hinge in adulthood and by the presence of a multivincular ligament with slender, narrowly and equally spaced resilifers (Muster 1995; Tëmkin 2006) and most likely polyphyletic (Tëmkin 2006), displays a large variety of shell shapes and strategies of byssal attachment (Printrakoon and Tëmkin 2008). Several species of Isognomon acquired adaptations to life on soft substrates. Large, thin-shelled modern species, e.g., the “mudflat Isognomon” of Printrakoon and Tëmkin (2008), may form epibyssate clusters that are all fixed to a single, horizontally aligned individual of Isognomon. A similar life style may be discussed for Isognomon rectangularis that occurs sympatric with Isognomon lusitanicus at a few localities of the Arranhó II member in the central Lusitanian Basin. In contrast, Isognomon rugosus and its descendant I. lusitanicus had an endobyssate life style, fastened to bioclasts within soft sediment (Fürsich 1980), similar to several species of Isognomon (Hippochaeta) from the Cainozoic of Europe (“orthothetic edgewise recliners”; Seilacher 1984). However, Isognomon (Hippochaeta) maxillatus (Lamarck, 1801) from the Mediterranean Pliocene likely achieved stabilisation almost entirely by weighting of the hinge area and by its byssus (Savazzi 1995); it lacks the broadened anterior shell surface that is well-developed in most specimens of Isognomon rugosus and Isognomon lusitanicus and helped them to attain a stable position lying on the sediment surface. Consequently, in the latter two species only the umbonal area may have penetrated the sediment, as suggested by the strongly forward projecting umbones commonly seen in I. rugosus and I. lusitanicus and by the gradual realignment of the ligament axis for an angle of up to 30° in gerontic individuals (Fig. 4C).

Unlike Arcomytilus morrisii and particular modern Isognomon species, Isognomon rugosus and Isognomon lusitanicus did not form clusters, but rather occurred in dense banks composed of live specimens of various growth stages but also containing dead shells. These banks may have extended for hundreds of square meters, as can be observed in the Sobral member at the Atlantic coast south of Santa Cruz (Fürsich et al. 2009). Large variability in shell shape and the development of strongly triangular morphs may also have been triggered by competition for food and related upward growth as inferred for Arcomytilus.

The multivincular ligament of Isognomon is composed of fibrous sublayers that fill the resilifers and lamellar sublayers that are placed between the resilia and connect the fibrous parts (Tëmkin 2006). In some samples from Portugal, part of the fibrous material is still preserved. Measurements of several specimens show that ligamental planes of opposing valves form a maximum angle of ∼50° in the central part of the ligament length, while angles are much narrower (∼30°) at the anterior and posterior portions of the ligament. These values are constant throughout ontogeny, which must have been achieved by constant shell thickening in the sub-ligamental region. Evaluating the thickness of the lamellar ligament portions, the valves may have been gaping with a maximum angle of ∼25° in live specimens. Whether less numerous but broader resilifers provided an adaptive advantage, e.g., with regard to strength or physiological costs, cannot be clarified based on fossil material.

The slight decrease of shell elongation that occurs in Eomiodon through time (Fig. 14) may be related to size increase. As can be inferred from shell characters, Eomiodon is a shallow-infaunal filter feeder. Consequently, increase in size and respective body mass may have caused a relative decrease in muscle strength of these bivalves (Stanley 1970, 1972). As a result, shortened shells, which would have had to burrow less deeply to attain their growth position with the posterior end just below the sediment-water interface, may represent an advantageous adaptation in large Eomiodon.

Ecophenotypy

It can be generally stated that ecophenotypy between associations or populations can not be proven based on the samples analysed in this study. On one hand, even among the exceptionally large dataset used for this study, many samples are too small to be statistically significant, and thus do not allow the use of more sophisticated statistical methods. On the other hand, intraspecific shape variability seen in all three taxa usually occurs within populations and thus may be driven by synecological factors. This pattern becomes obvious in most of the population plots (SOM: figs. S1–S10), as small samples are commonly included in the morphospace of the largest samples from a particular time slice or lithostratigraphic unit.

In Arcomytilus the sampled populations may be assigned to nine different associations or assemblages representing euhaline to mesohaline conditions. However, no distinct shape modification occurs, that could be linked to salinity or faunal composition. A single relatively small population, i.e., [6M] from the Arranhó II member [6] that is assigned to the autochthonous Arcomytilus morrisii/Corbulomima suprajurensis assemblage, stands out, due to its extremely small shape variability (SOM: fig. S3). Unfortunately, no other samples from this widespread assemblage are available, as well-preserved specimens from the more or less unconsolidated mud-facies are extremely hard to obtain. Two additional samples that occupy a similar, small morphospace [5F, 6Q] come from the Choffat collection and cannot be assigned to a particular ecological group. The same is true for samples [4R] and [4T] that include somewhat aberrant, more straight and slender specimens than the other populations from the Sobral member [4] (SOM: fig. S2).

To a certain extent, the high variability seen in Arcomytilus from the Arranhó I [5] and Arranhó II [6] members may reflect ecophenotypy. In the aforementioned strata, the preferred habitat of Arcomytilus morrisii were soft or even soupy sediments in shallow lagoonal settings, either siliciclastic or calcareous in nature. All populations from these soft bottom settings include a major portion of moderately to strongly curved and/or broadened, triangular shells, while the contemporaneous more sandy, less soupy facies of the Sobral member [4] were inhabited by less variable populations of common shape. As already outlined, broad, triangular shells may be advantageous in calm water habitats (Akester and Martel 2000). Moreover, the increase of shell surface in these broadened specimens may have rendered a purely epibyssate mode of life rather inconvenient. As a result, a semi-infaunal life style has been adapted in the respective settings, as indicated by specimens with tapering, curved beaks.

Results from rib density analysis show clearly that an adequate temporal and spatial sample density is indispensable for disentangling true evolutionary trends and ecophenotypic plasticity. An evolutionary trend towards fewer ribs is suggested by the data from all units in the Lusitanian Basin—with exception of the specimens of Arcomytilus from the Sobral member [4] (Fig. 10). In the samples from this member, an increase in rib numbers argues against the general trend. Moreover, rib numbers are obviously not correlated with adult shell size, as many small adults from, e.g., the Cabo Mondego and Alcobaça formations tend to have even more ribs as numerous large shells from younger strata. Ribs may bifurcate or additional ribs may become inserted at any growth stage. Commonly, these processes occur periodically, but often also punctually even on a single shell, without any identifiable fixed pattern. This renders a substantial effect of numbers and strength of ribs on shell stability unlikely. Moreover, shell shape has no influence on rib density (Fig. 12). However, there must be an adaptive value of high, respectively low, numbers of ribs. Potentially, rib density may be linked to the substrate. Less numerous but somewhat more prominent ribs may be advantageous in soupy sediments (carbonate mud in Arranhó I [5] or silty clay in Arranhó II [6] members), possibly to stabilise the individuals in the sediment, while the specimens settling on the silty to fine-sandy sediments of the Sobral Member may have benefited from more numerous, less elevated ribs in any way. Whether less numerous ribs represent a hydrodynamic advantage in quiet settings, and vice versa, remains to be tested. This is, however, beyond the scope of this study.

Ecophenotypy in Isognomon rugosus can clearly be negated. All larger samples show similar, modest variability with regard to shell shape and more or less produced umbones (SOM: figs. S4, S5), although they come from two assemblages and four different associations, one of them with four distinct subsets. As a result, shell form is influenced by individual competition for space and food and likely by smallscale variations in substrate rather than by any other environmental parameter. A similar degree of variability is displayed by Isognomon lusitanicus that come from four associations and three subsets (Table 1). Occasionally, single populations, though showing large areas of overlap in their convex hulls, may develop slightly different gross shapes (SOM: figs. S6, S7). However, these variations cannot be correlated with ecological features with certainty and samples from different associations may well fill a similar morphospace, while samples from the same subset may occupy slightly disparate positions in the plots. For example, all populations from localities of Arranhó I [5] and Arranhó II [6] members at the Atlantic coast tend to include more triangular shells with pointed beaks, while rectangular forms with short beaks dominate in populations from the Arruda subbasin. However, all of these populations overlap for a large part of their morphospace and there is no evidence from these data that supports the existence of geographic subspecies in statu nascendi, although geographical isolation in the subbasins may have occurred. As similar geographically isolated and ecologically stressed environments are known to cause frequent rapid speciation (e.g., Müller et al. 1999; Harzhauser and Mandic 2008), at least some attention should be paid to this hypothesis.

In Eomiodon, a significant correlation between shell shape and environmental conditions could not be detected. Moreover, shape variability is generally low within certain time-slices and lithostratigraphic units. From Fig. 14B and SOM: fig. S8, it becomes obvious that relatively few individuals are responsible for the greater morphospace occupied by Eomiodon from the Cabo Mondego formation [2]. The two large populations from the Alcobaça formation [3] show almost equal variability and largely cover the shape variability of smaller samples. However, no palaeoecological data exist for the large samples. Slight differences in shape variability among populations can be inferred from the Sobral member [4] (SOM: fig. S9). However, populations from both the Eomiodon securiformis association and the Eomiodon securiformis/nerineid sp. B association include specimens of similar shape within and between populations (SOM: fig. S9) and differences between samples from similar or different associations are more or less equal. Hence, there is no indication for ecophenotypy in Eomiodon.