Research Article |
Corresponding author: Paul C. Southgate ( psouthgate@usc.edu.au ) Academic editor: Frank Köhler
© 2023 Paul C. Southgate, Thane A. Militz.
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Citation:
Southgate PC, Militz TA (2023) A multivariate approach to morphological study of shell form in cowries (Gastropoda, Cypraeidae): a case study with Umbilia armeniaca (Verco, 1912). ZooKeys 1158: 69-89. https://doi.org/10.3897/zookeys.1158.98868
|
Multivariate approaches to morphological study of shell form have rarely been applied to cowries (Gastropoda: Cypraeidae) with preference, instead, for comparing formulaic notations of shell form that report averages (i.e., means) for key morphometrics such as shell dimensions, their ratios, and counts of apertural teeth. Although widely applied, the “shell formula” does not account for variation among individuals or support statistical comparison between taxa. This study applied a multivariate approach to analyse shell form within the four accepted subspecies of the cowrie, Umbilia armeniaca (Verco, 1912) and included a previously unstudied, and most northerly, population of U. armeniaca from Lancelin, Western Australia. Multivariate analyses readily separated the recognised subspecies of U. armeniaca (U. a. armeniaca, U. a. diprotodon, U. a. clarksoni and U. a. andreyi), but did not separate the Lancelin population from U. a. andreyi, indicating that the former represents a northward extension of U. a. andreyi that is not morphometrically distinguishable. These results provide improved understanding of infraspecific differences in shell form of U. armeniaca across its broad distribution, and demonstrate the utility of multivariate morphometric methods for statistical comparison of shell form between taxa. This approach is complimentary to existing research practices and has broad potential application in future morphometric studies of both extant and fossil taxa within the family Cypraeidae.
Cowry, gastropod, marine, morphometrics, shell form, taxonomy, Umbilia
The family Cypraeidae Rafinesque, 1815 (cowries) comprises a large group of marine gastropods characterised by colourful, generally glossy, shells with a narrow, elongate aperture bordered by teeth. Cowries demonstrate variable inter- and infraspecific shell morphology (
Despite broad application of molecular approaches in modern gastropod taxonomy (e.g.,
Application of statistical tests to compliment comparisons of shell form between and among groups of cowries can, in its simplest form, assess differences for each morphometric separately (e.g., univariate analysis). When considering a morphometric in this manner (e.g.,
To demonstrate how a multivariate approach to morphological study of shell form might benefit research into cowrie systematics, this study validates existing infraspecific taxonomy for the Australian endemic Umbilia armeniaca (Verco, 1912). Specific conchological attributes are attributable to location (
The cowrie genus Umbilia Jousseaume, 1884 is represented by five living species endemic to Australia. They have limited larval dispersal because of intracapsular development (
Four subspecies of Umbilia armeniaca are currently recognised (
Specimens of the four recognised subspecies of Umbilia armeniaca A U. armeniaca armeniaca, trawled off Ceduna, Great Australian Bight, 90–120 m, 105 mm B U. armeniaca diprotodon, taken by diver, Thorny Passage, Port Lincoln, South Australia, 35 m, 102 mm C U. armeniaca clarksoni, taken by diver off Cape Le Grande, Esperance, Western Australia, 30–35 m, 94.1 mm D U. armeniaca andreyi, collected using ROV, off Augusta, Western Australia, 150 m, 84.2 mm and E U. armeniaca from the Lancelin population, collected using ROV, off Lancelin, Western Australia, 200 m, 68.9 mm.
Primary data for shell length (L), shell width (W), and shell height (H), columellar (CT) and labral (LT) tooth counts, and shell mass (M) were collected from previously unstudied specimens of U. a. armeniaca (n = 21), U. a. diprotodon (n = 1), U. a. clarksoni (n = 2) and U. a. andreyi (n = 4), following the methodology of
Diagrammatic representation of how a multivariate approach to morphological study of shell form for cowries was applied in this study, using Umbilia armeniaca as an example (see Suppl. material
All data analyses were performed using R (version: 4.2.1), an open-access software environment for statistical testing and graphics, with the stats (
The morphometrics considered in this study were those proposed by
Because both dimensionless (e.g., ratios) and differently scaled (e.g., length vs. tooth counts) morphometrics were considered, values were transformed to Z-scores prior to ordination and statistical testing (R function: scale) (Fig.
Using the morphometric Z-scores for uncensored specimens of U. a. armeniaca (n = 51), U. a. diprotodon (n = 28), U. a. clarksoni (n = 16), U. a. andreyi (n = 18), and the Lancelin population (n = 17), a resemblance matrix was computed based on Euclidean distances between specimens (R function: vegedist) (Fig.
The influence of each morphometric on the patterns visualised with nMDS (Fig.
Despite numerous advantages of ordination, visual interpretations of multidimensional data after reducing dimensionality can be subjective (
To contextualise how the multivariate approach compared with a univariate approach, an analysis of variance (ANOVA), constructed as a linear model, was used to test whether the means of each morphometric differed among the five U. armeniaca groups examined (R function: lm). Pairwise comparisons between groups were made using linear hypothesis tests of the estimated marginal means (R function: emmeans), controlling for the family-wise error rate with the
Among the studied specimens of Umbilia armeniaca, the a priori assigned groups (i.e., U. a. armeniaca, U. a. diprotodon, U. a. clarksoni, U. a. andreyi, and the Lancelin population) were able to explain a significant amount (R2 = 0.48, F = 29.32, P < 0.001) of the variation in shell form (Fig.
A nMDS ordination (stress = 0.16) of the resemblance matrix for Umbilia armeniaca, where shaded ellipses indicate the 95% confidence interval of group (subspecies or population) centroids and plot characters indicate data source B–H Associations between ordination structure and morphometrics influencing this structure, where the green lines illustrate B length C height:length ratio D width:length ratio E height:width ratio F normalised columellar tooth count G normalised labral tooth count, and H relative mass contour lines.
Results of pairwise comparisons testing the hypotheses that there were no differences in central tendency (i.e., centroid) of shell form among the studied Umbilia armeniaca groups (subspecies or population). The Euclidean distance (D) between centroids, coefficient of determination (R2), and Holm-adjusted probability that the distance between centroids arose by random chance (P) are presented.
U. armeniaca group | andreyi | armeniaca | clarksoni | diprotodon | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
D | R2 | P | D | R2 | P | D | R2 | P | D | R2 | P | |
armeniaca | 2.86 | 0.29 | 0.001 | – | – | – | – | – | – | – | – | – |
clarksoni | 4.66 | 0.64 | 0.001 | 2.48 | 0.23 | 0.001 | – | – | – | – | – | – |
diprotodon | 3.20 | 0.47 | 0.001 | 1.61 | 0.14 | 0.001 | 2.70 | 0.40 | 0.001 | – | – | – |
Lancelin population | 1.00 | 0.09 | 0.012 | 2.89 | 0.31 | 0.001 | 4.60 | 0.71 | 0.001 | 3.41 | 0.56 | 0.001 |
Results also indicated that a similar degree of variation (i.e., dispersion) in shell form existed for the U. armeniaca subspecies (Fig.
Results of pairwise comparisons testing the hypotheses that there were no differences in variation (i.e., dispersion) in shell form among the studied Umbilia armeniaca groups (subspecies or population). The mean (x̄) ± standard deviation (SD) and range in Euclidean distance that specimens were from their group centroid are presented. Shared alphabetic superscripts identify group means that are not significantly (Holm-adjusted P ≥ 0.01) different.
U. armeniaca group | distance from centroid* | |
---|---|---|
(x̄ ± SD) | range | |
andreyi | 1.72 ± 0.53ab | 0.88 – 3.29 |
armeniaca | 1.94 ± 0.56a | 0.87 – 3.24 |
clarksoni | 1.53 ± 0.61ab | 0.73 – 2.91 |
diprotodon | 1.50 ± 0.49ab | 0.63 – 2.59 |
Lancelin population | 1.22 ± 0.46b | 0.48 – 2.20 |
All morphometrics considered representative of shell form (i.e., L, H/L, W/L, H/W, nCT, nLT, mR) significantly influenced the ordination structure of the U. armeniaca groups visualised in Fig.
By examining each morphometric independently, relative differences in morphometric values among groups inferred from the nMDS ordination (Fig.
Box plots showing univariate comparisons of A shell length B height:length ratio C width:length ratio D height:width ratio E normalised columellar tooth count F normalised labral tooth count, and G relative mass among the studied Umbilia armeniaca groups (subspecies or population). Black diamonds represent group means, boxes illustrate first and third quartile as box edges and median as central line. Shared alphabetic superscripts identify group means that are not statistically different (Holm-adjusted P ≥ 0.01) for a morphometric.
Considering that not all morphometrics were consistently similar or dissimilar between groups, it was not possible to conclude whether groups differed in overall shell form from univariate comparisons alone. This conundrum is best illustrated by a comparison of U. a. andreyi and the Lancelin population, where specimens from these groups differed in central tendencies of relative mass (Fig.
Our results confirm that variation in shell form is prominent within and between populations of Umbilia armeniaca. Such variability may represent ecophenotypic responses in shell form to environmental factors, random genetic variations independent of adaptive value, or natural selection. Results of this study cannot directly distinguish between these, or other possible causal mechanisms promoting variation in the shell form of cowries, within or between populations, which have seen much discussion elsewhere (see
Crucial to the validity of taxa differentiated through morphological study of shell form is a replicable and objective approach for comparison. Prior morphological study of shell form in U. armeniaca has relied on subjective comparisons of morphometrics and their central tendencies to resolve infraspecific differences, with statistical testing limited to a comparison of relative mass (
In their original description of U. armeniaca subspecies,
Certainly, multivariate distributions (Fig.
When comparing taxa, it is important to recognise that not all diagnostic factors can be reliably incorporated into multivariate analysis. Shell pattern, for example, is important in cowrie characterisation (
Of the four U. armeniaca subspecies, the nMDS ordination revealed that shells of the previously unstudied Lancelin population were most similar to the neighbouring population of U. a. andreyi. Although Lancelin shells had a significantly reduced relative mass, compared to U. a. andreyi, this difference was insufficient to differentiate the Lancelin population from U. a. andreyi when accounting for the overall variability in shell form. Differences in relative mass are closely associated with differences in shell callosity (
Most datasets used in the study of shell form are multidimensional and, consequently, a large component of any systematic study will involve consideration of how to extract a meaningful summary of differences in shell form among specimens and/or groups. The multivariate approach taken in this study, couples easy-to-interpret graphics produced via ordination with an objective appraisal of inter-group differences via statistical testing. When combined with a univariate approach, as done here, a broad range of questions of relevance to systematics can be addressed objectively. For example, if the goal is to characterise differences in shell form between groups, multivariate tests comparing differences in central tendencies (such as group centroids) or variability (such as group dispersion) in multidimensional space are most appropriate. Furthermore, if the goal is to characterise how a particular morphometric differs between groups, univariate tests comparing differences in central tendencies (such as group means) for that morphometric will be of value. Regardless of the tests used, statistical testing should only be employed to address questions framed within an appropriate statistical context if sample sizes are large enough to permit detection of statistically significant differences (
It is also important to consider which morphometrics are most appropriate for statistical testing. Studies comparing shell form among cowries have, over time, varied greatly in the morphometrics selected to represent shell form (
Aside from representing the primary sources of infra- and interspecific variation in shell form of cowries (
Multivariate approaches to the morphological study of shell form of cowries have been utilised primarily to develop hypotheses related to ecological and functional diversity within the target species (
This study has demonstrated the utility of a multivariate approach that couples easy-to-interpret graphics produced via ordination with an objective appraisal of inter-group differences via statistical testing with clearly defined thresholds for both outlier detection (i.e., |Z-score| > 3) and resolving infraspecific differences (P < 0.01). Using Umbilia armeniaca as a case study, we showed how primary data from unstudied specimens might supplement secondary data from prior studies to validate existing infraspecific taxonomy and characterise a previously unstudied (Lancelin) population of this species. The multivariate approach showed the four recognised U. armeniaca subspecies to be similarly variable, but confirmed differences in central tendency of shell form. Our analysis did not justify differentiation of the Lancelin population of U. armeniaca which is best considered a northward extension of U. a. andreyi. Results of this study provide improved understanding of intraspecific differences in shell form of U. armeniaca across its broad distribution, and demonstrate how multivariate morphometric methods for statistical comparison of shell form between taxa might benefit cowrie systematics. This approach is complimentary to existing research practices and has broad potential application in future morphometric based studies of cowries.
We thank Mr Ray Walker of Busselton, Western Australia for providing the Lancelin specimens used in this study and Nittya S.M. Simard for assistance with figure compilation. This study was supported by University of the Sunshine Coast Research Initiative funding to the first author.
Measures of central tendency (i.e., mean and median) and variation (i.e., standard deviation and range) among the studied Umbilia armeniaca groups (subspecies or population) for each morphometric considered representative of shell form. Shared alphabetic supercrips identify group means that are not statistically different (Holm-adjusted P ≥ 0.01) for a morphometric.
Morphometric | Umbilia armeniaca groups | ||||
---|---|---|---|---|---|
andreyi | armeniaca | clarksoni | diprotodon | Lancelin | |
length (mm) | |||||
x̄ ± SD | 73.3 ± 5.8c | 94.9 ± 11.1b | 92.9 ± 5.4b | 107.5 ± 4.5a | 70.2 ± 2.5c |
median | 71.3 | 94.2 | 93.9 | 107.0 | 69.5 |
range | 65.5–84.0 | 73.9–118.1 | 79.8–98.8 | 100.7–117.5 | 66.3–75.0 |
height:length ratio | |||||
x̄ ± SD | 0.563 ± 0.016a | 0.547 ± 0.018b | 0.504 ± 0.020d | 0.533 ± 0.017c | 0.569 ± 0.013a |
median | 0.566 | 0.545 | 0.502 | 0.531 | 0.570 |
range | 0.530–0.584 | 0.505–0.592 | 0.472–0.553 | 0.500–0.569 | 0.541–0.588 |
width:length ratio | |||||
x̄ ± SD | 0.633 ± 0.017ab | 0.628 ± 0.023ab | 0.602 ± 0.019c | 0.623 ± 0.019b | 0.642 ± 0.017a |
median | 0.632 | 0.622 | 0.596 | 0.619 | 0.646 |
range | 0.604–0.668 | 0.590–0.687 | 0.579–0.649 | 0.589–0.664 | 0.606–0.666 |
height:width ratio | |||||
x̄ ± SD | 0.889 ± 0.017a | 0.872 ± 0.020b | 0.837 ± 0.012d | 0.854 ± 0.017c | 0.885 ± 0.014a |
median | 0.889 | 0.874 | 0.838 | 0.857 | 0.884 |
range | 0.850–0.921 | 0.805–0.853 | 0.815–0.853 | 0.826–0.901 | 0.869–0.911 |
normalised columellar teeth | |||||
x̄ ± SD | 16.7 ± 1.2b | 18.4 ± 1.1a | 18.6 ± 1.0a | 18.3 ± 1.0a | 16.3 ± 0.5b |
median | 16.7 | 18.4 | 18.4 | 18.5 | 16.2 |
range | 14.2–19.4 | 15.1–20.3 | 17.3–20.2 | 16.2–20.5 | 15.4–17.1 |
normalised labral teeth | |||||
x̄ ± SD | 21.6 ± 1.0b | 23.2 ± 1.2a | 23.3 ± 1.3a | 21.8 ± 0.9b | 21.4 ± 0.9b |
median | 21.8 | 23.4 | 23.0 | 21.9 | 21.6 |
range | 19.7–23.4 | 20.8–25.4 | 21.5–26.3 | 19.5–23.4 | 19.6–23.9 |
relative mass | |||||
x̄ ± SD | 13.7 ± 1.7a | 10.7 ± 1.2c | 8.5 ± 0.7d | 11.8 ± 0.9b | 12.2 ± 1.1b |
median | 14.0 | 10.6 | 8.3 | 11.8 | 12.2 |
range | 10.6–16.6 | 8.6–13.9 | 7.2–9.6 | 10.3–13.6 | 10.5–14.3 |
Annotated code pertaining to the multivariate approach
Data type: docx file