Characterization and multivariate analysis of Patella intermedia, Patella ulyssiponensis and Patella vulgata from Povoa de Varzim (Northwest Portugal)

17 Pages·2003·7.3 MB·English

by  Joao Paulo Cabral

© SociedadEspañola deMalacologia Iherus,2\ (2): 1-17, 2003
Characterization and multivariate analysis oíPatella inter-
media^ Patella ulyssiponensis and Patella vulgata from Póvoa
de Varzim (Northwest Portugal)
Caracterización y análisis multivariante de Patella intermedia^ Patella
ulyssiponensis and Patella vulgata de Póvoa de Varzim (noroeste de
Portugal)
Joáo Paulo CABRAL*
Recibidoel9-IX-2002. Aceptado el13-1-2003
ABSTRACT
Samples of Patella intermedia, Patella ulyssiponensisand Patella vulgatawere collected from
the lowestto the highest level ofA Ver-o-Mor and Agucadoura beaches (Póvoa de Varzim,
Portugal). Identification ofthe specimens was based on the morphology ofthe radula pluri-
cuspid teeth and of the shell. Regressions of shell width or shell apexpos vs. shell length
yielded slopes not statistically different, but slopes of log shell height vs. shell length regres-
sion lines were significantly different, indicating thatthese limpets have different shell grow-
ing patterns. However, canonical discriminantanalyses using onlyvariables describing shell
form yielded poor discrimination between species. Analysis using onlyvariables describing
radula relative size improved discriminaron, wasvery satisfactoryfor P. ulyssiponensis spec-
imens, but resulted in a low identification of P. intermedia. The bestresultswere achieved us-
ing both shell form and radula relative size variables, and the most discriminating variable
was radula length/shell height ratio, instead of radula length/shell length as reported in the
literature. fHowever, whereas correct identification of P. ulyssiponensis specimens was very
good (higher than 95 %), for P intermedia and P vulgata, this valué ranged from 70 - 80
%. In light of data reported in the literature, morphometric characteristics such as those uti-
lized in this work are probably useful for characterization of P ulyssiponensis as a species,
but are of limited valué for P intermedia and P vulgata. Qualitative characters are thus still'
indispensable for the discrimination between these two Patella species. Very good qualita-
tive discriminators were, for P intermedia, the tall and broad cusp 2 and the dark marginal
rays in the shell interior surface, and for P vulgata, the pointed protuberance on cusp 3, and
the silvery head scar.
RESUMEN
Muestrasde Patellaintermedia, Patella ulyssiponensisy Patella vulgatafueron colectadasen
una franja abarcando desde el nivel mas bajo al mas alto de las playas de A Ver-o-Mar y
Agucadoura (Póvoa de Varzim, Portugal). La identificación de los especimenes se basó en
la morfología de la concha y de los dientes pluricúspides de la ródula. Las pendientes de
las regresiones entre la anchura o el apexpos (distancia entre la proyección del ápice y la
parte posterior de la concha) de la concha y la longitud de la concha no difirieron estadís-
*UniversityofOporto,CenterofMarineandEnvironmentalResearch(CIIMAR)andFacultyofSciences,Rúa
doCampoAlegre, 1191,4150-181 Porto, Portugal,[email protected]
Iberus, 21 (2), 2003
ticamente, mientras que las pendientes de las regresiones entre el logaritmo de la altura y
la longitud de la concha fueron significativamentediferentes, indicandoqueestas lapas pre-
sentan patrones de crecimiento distintos. Sin embargo, el análisis canónico discriminatorio
utilizando solo lasvariablesquedescriben la forma de la concha tuvo un bajo poderdedis-
criminación entreespecies. Al utilizarsolo lasvariables quedescriben el tamaño relativode
la ródula se mejoró el poder discriminatorio del análisis, y aunque permitió una identifica-
ción satisfactoria de P. ulyssiponensis, pero resultó tener bajo poder para la identificación
deP. intermedia. Losmejoresresultadosseobtuvieron utilizandovariablesrelacionadastanto
con laformadelaconchacomoconeltamañorelativode la ródula, ylavariablecon mayor
poder discriminatorio fue la razón longitud de la ródula/altura de la concha, en lugar de
longitud de la ródula/longitud de laconcha, comose ha reportadoen la bibliografía. Mien-
tras que la correcta identificación de especímenes de P. ulyssiponensis fue muy alta (mayor
que 95 %), para P intermediay P vulgata, estevalorvarió entre 70- 80 %. A la luz de los
datos publicados en la bibliografía, características morfométricas tales como las utilizadas
en el presente trabajo, pueden ser de gran utilidad para la caracterización específica de P
ulyssiponensis, pero tienen un valor limitado para la caracterización de P intermedia y P
vulgata. Lascaracterísticascualitativas, portanto, siguen siendo indispensables para la dis-
criminación ente estas dos especies de Patella. Para P intermedia, \a segunda cúspide alta
yancha, asícomo los radios marginalesoscurosen la superficie interna de la concha, cons-
tituyeron muy buenos discriminadores cualitativos, mientras que para P vulgata, lo fueron
la protuberancia puntiaguda en la tercera cúspide, junto con la cicatriz de la cabeza de
color plateada.
KEYWORDS: Patella, radula,canonicaldiscriminantanalysis, morphometry.
PALABRAS CLAVE:Patella, rádula,anñalisiscanónicodiscriminante, morfometría.
INTRODUCTION
The characterization and delimita- theAlgarve (Portugal) forP. vulgata, and
tion of Patella intermedia Murray in northern África for P. intermedia. P.
Knapp 1857, Patella ulyssiponensis ulyssiponensis is the species with the
Gmelin 1791, and Patella vulgata Linné least variable shell (Fischer-Piette and
1758, has been a source of debate and Gaillard, 1959;Fisher-Piette, 1966).
controversysincetheywereproposedas ThedifficultiesintheIdentificationof
species (FiSHER-PiETTE, 1935; Rampal, P intermedia, P. ulyssiponensis and P.
1965; IBAÑEZ, 1982). This was due to the vulgata specimens based on shell mor-
high shell variability commonly exhib- phology alone led to the search for new
ited by these species. In regions of low characters. Dautzenberg, in the end of
variability, P. intermedia, P. ulyssiponensis the 18th century, pointed out the impor-
and P. vulgata shells are usually distinct tance ofradula length and pignientation
from each other. In regions ofhigh vari- ofmantletentacletodiscrimínateseveral
ability, however, some shells display Patella species, being therefore the first
intermedíatecharacteristics, and reliable malacologist topropose alternative char-
Identification of these species based on acteristics to shell morphology (Fischer-
shell morphology alone is difficult Piette and Gaillard 1959; Rampal,
(FiSHER-PiETTE, 1934, 1948, 1966; Evans, 1965). The work of Fisher-Piette (1934,
1947, 1958; Fisher-Piette and Gail- 1935) showed, for the first time, that the
LARD, 1959). Shell variabiHty of P. morphology of the radula pluricuspid
vulgata andP. intermedia is highest in the teeth could be used to characterize
Basque coast of France and Spain, and severalPatella species, sincethey usually
in the south limit of their distribution. display low intraspecific variability, and
CabraL: Characterization and multivariate analysis oí Patellaspp.in Northwest Portugal
Figure 1. Location ofsamplingsites (AVer-o-MarandAgu^adourabeaches) nearPóvoadeVarzim
Inset: Localization ofthe town ofPóvoadeVarzim and Oporto districtin Portugal.
Figura 1. Ubicación delazonademaestreo (playasdeA Ver-o-MaryAgugadoura) en lascercaníasde
PóvoadeVarzim. Recuadro: UbicacióndelpueblodePóvoadeVarzimyeldistritodeOportoenPortugal
marked interspecific differences. Fisher- morphometric characters describing the
Piette initial observations were con- radula, shell,and softparts. Inparticular
firmed by EsLiCK (1940) and Evans thefollowingquestionswere.raised:
How
(1947). In a latter paper, Fisher-Piette - variable are these species in
ANDGaillard(1959)showedthatradula Póvoa deVarzim?
unicuspid teeth could also be used to - Is shell form similar in these
characterizeseveralPatellaspecies. limpets?
Other characteristics have been pro- - Are radula and shell characteristics
posed for specific discrimination in the equally important for the separation of
genus Patella: the foot colour, the breed- thesespecies?
ing seasons, and the ratio of the radula - Can these species be discriminated
length / cubic root of the shell volume by morphometric characters alone,
(FiSHER-PiETTE, 1935, 1941, 1948; Evans, rather than by morphological character-
1947, 1953, 1958; Fisher-Piette and istics?
Gaillard, 1959; Rampal, 1965; Powell,
1973).
Inthisstudy,samplesofP. intermedia, MATERIALAND METHODS
P. ulyssiponensis and P. vulgata from
Póvoa de Varzim (northwest Portugal) Collections were made at A Ver-o-
were examined and compared with Mar and Agu^adoura beaches, two very
respect to several morphological and similar and exposed shores situated
Iberus, 21 (2), 2003
—
\X^;J,u < Height >
"\
y^
11 y^
r Apexante \.
^
' S
1
1
{
\
Apeícpos /
{ Length
Figure2. Shell measurementsused incanonicaldiscriminantanalysis.
Figura2. Medicionesdelaconchautilizadasenelanálisiscanónicodiscriminatorio.
near the town of Póvoa de Varzim (Fig. the animáis in each área removed from
1). The upper shore is composed of dis- the rock. The total number of collected
persed and heavily eroded graniteboul- animáiswas608.
ders. Barnacles {Chthamalus) cover most Specimens were analysed for foot,
ofthesurfacesandP. imlgata is the dom- radula and shell characteristics. In the
inant limpet. Some mussels and gas- laboratory,theanimáiswereobservedfor
tropods live in the crevices. The middle thefoot colour, and then immersed fora
shore is composed of horizontal granite few minutes inboiling water to sepárate
platforms, very eroded, with small the shell from the soft part. The radula
sandy beaches in between. P. intermedia was removed from the visceral mass by
and P. inilgata are common in middle dissection, immersed in household
shore. P. ulyssiponensis is restricted to bleach to remove mucilaginous sub-
lower shore levéis, where P. intermedia stances, washed in distilled water, and
and P. vulgata also occur. Most of the P. measured to the nearest Imm using a
idyssipoiiensis shells are covered with ruler. After air-drying, pluricuspid teeth
abundant Gelidiiim pidchelum (Turner) werefirstobservedusingabinocularmi-
Kützing. Common macroalgae in the croscope, with 80 x final magnification.
middle and lower shore include Bifiir- Structural details of the teeth were ob-
caria bifiircata R. Ross, Chondrus crispiis served by scanning electrón microscopy,
Stackhouse, Codium tomentosiim (Hud- in selected specimens. The samples were
son) Stackhouse, coralline rhodophytes, gold coated (in a JEOLJFCllOO model;
Cystoseira sp., Enteromorpha spp., Gi- film thickness less then 20 nm), and ob-
gartina pistillata (Gmelin) Stackhouse, served in a JEOLJSM-35C model scan-
Gracilaria vernicosa (Hudson) Papenfuss, ning electrón microscope, working at 15
mm
Lythophylhim spp., Mastocarpiis stellatns keV,with39 workingdistance.
(Stackhouse in Withering) Guiry and The external and internal shell sur-
Sargassmn muticiim (Yendo) Feusholt. faces were examined and their charac-
Sampling tookplace inJanuary and No- teristics were recorded. Shell length,
vember2000,atlowtide. width, height, apexante and apexpos
Specimens were collected from the (Fig. 2 and Table I for the definition of
lowest to the highest level of the shores. these measures) were then determined
Squared áreas with ca. 30 x 30 cm were to the nearest 0.01mm using a digital
marked at random in each level, and all calliper (Mitutoyo, modelCD-15DC).
Cabral: Characterization and multivariateanalysis o^Patellaspp.in Northwest Portugal
Table 1. Characters usedin thecanonical discriminantanalysis.
TablaI. Caracteresusadosenelanálisiscanónicodiscriminante.
Acronym Description
SL Shelllength:greatestdistancebetweenanteriorandposteriorends
SW Shellwidth:greatestdistanceperpendiculartotheanterior-posterioroxis
SH Shellheight:greatestverticaldistancefromapexoftheshelltotheplañeofaperture
SAA Shellapexante:greatestdistancebetweenapexandanteriorend
SAP Shellapexpos:greatestdistancebetweenapexandposteriorend
SV Shellvolume= (tt/S) x (SW/2) x (SL/2) xSH
RL Radulalength
Morphometric analysis of the shell liability of previous findings (using a
and radula was carried out using the posterioriprobabilities). For each species,
variables commonly utilized in studies specimens were re-ordered randomly
of patellogastropoda (Ibañez, 1982; using Excel 2000 program RAND func-
Ibañez and Feliu, 1983; Feliu and tion. Each group (species) was then di-
Ibañez, 1984; Hernández-Dorta, 1992; vided in two subgroups. For P. interme-
SiMisoNANDLindberg, 1999). Themean dia, these contained 152 specimens each,
and coefficient of variation of the mean forP. ulyssiponensis, 33 and 32, each, and
was calculated for each variable and for P. vidgata, 119 and 120. The first
species. Linear regression analysis using group (the analysis sample containing
the least squares method was applied to the training or calibration data) was
pairs of shell or radula variables. These used to derive the discriminant func-
calculations were carried out using tions, coefficients, and loadings. The
Microsoft® Excel 2000 program. Com- second group (thehouldoutsamplecon-
parison ofthe slopes was carried outby taining the test data) was used to test
analysis of covariance using the statisti- the discriminant functions, by classify-
calproceduredescribedbyZar (1984). ing each specimen into one ofthePatella
Canonical discriminant analysis on species. This process was repeated ten
shell and radular morphometric vari- times. At each run, a percent correct
ables was used to reveal differences Identification to species was evaluated.
among the populations, and to identify The means of these valúes were com-
the variables that were responsible for pared with those obtained by using a
the majority of separation between posteriori probabilities. Discriminant
species. Discriminant analyses were car- analyses were performed using the
ried out in several steps. Firstly, using a NCSS (NCSS Statistical Software, Utah,
posteriori probabilities and no selection USA) and XLSTAT 5.1 (Addinsoft, Paris,
of variables, four discriminant analyses France) softwarepackages.
were carried out, two using only shell
characters, one with only radula charac-
ters, and one using simultaneously shell RESULTS
and radula characters. Secondly, using a
posteriori probabilities, and simultane- I. Delimitationofspecies
ously shell and radula characters, dis-
criminant analysis was carried out with Based on radularand shellmorphol-
stepwise selection of variables. Signifi- ogy, the collected specimens were
cance level "to enter" and "to stay" the divided into three groups. Within each
variables was set at 0.15. Thirdly, cross- group, radula pluricuspid teeth dis-
validation analysis (usinga priori proba- played low variability, and shells exhib-
bilities) was undertaken to assess the re- ited limited morphologicalvariation.
Iberus, 21 (2), 2003
Figure3. Morphologyofpluricuspidteethoftheradula.Numberingofthecuspsbeginsfromleftto
theright(correspondingtotheouterandinnersidesoftheradula).Top:Patellaintermedia. Cusp2is
taller and broader than cusps 1 and 3. Middle: Patella ulyssiponensis. Cusp 2 is directed to cusp 3.
Cusp 3 iswitherthan cusp 2. Cusp 3 has aprotuberanceon its outerside. Bottom: Patella vulgata.
Cusps2and3subequal.Cusp3hasapointedprojectiononitsoutersidebordenScalebars 100}im.
Figura3.Morfologíadelosdientespluricúspidesdelaradula. Lanumeracióndelascúspidescomienzadesde
la izquierda hacia la derecha (correspondiendo a los lados externos einternosde la radula). Superior:
Patellaintermedia. Lasegundaciíspideesmásaltayanchaquelaprimeraytercera.Medio:Patellaulys-
siponensis. Lasegundacúspideestádirigidahacialatercera. Estaúltimaesmásanchaquelasegunda. La
terceracúspidetieneunaprotuberanciasobreelladoexterno. Inferior:Patellavulgata. Lassegundayter-
ceracúspidessonsubiguales. Laterceratieneunaproyecciónpuntiagudaenelbordeexterno. Escalas100fon.
The three groups displayed the fol- (left) and 3 (right). Margin of the shell
lowingfeatures: rimóse, with pointed extensions con-
Group L Three unequal pluricuspid nected to the rays (Figure 4). External
teeth (Figure 3, top). Cusp 2 (centre) surface with few and prominent ribs.
much taller and broader than cusps 1 Interior with alteniating dark and light
CabraL: Characterization and multivariate analysis o{I^itellaspp.in Northwest Portugal
Figure4. Patellaintermedia bhclls tromAVer-o-Mar andAguijjdüura bcaches.
Figura4. ConchasdePatellamizirat^VÁprovenientesdelasplayasdeA Ver-o-MaryAgugadoura.
rays,inthelowerpart. Headscaryellow- (Figure 5). External surface crowded by
orange,butcreamyinafewspecimens. numerous, closely spaced, well-marked
Group II. Three unequal pluricuspid ribs, of unequal size. Interior homoge-
teeth (Figure 3, middle). Cusp 1 (left) neous, porcelanous white. Head scar
very small. Cusp 2 (centre) bent to cusp white,creamyorpaléorange.
3 (right). Base ofcusp 3 widerthancusp Group III. Three unequal pluricuspid
2. Cusp 3 with a protuberance on its teeth (Figure 3, bottom). Cusp 1 (left)
outer side - a vestigial fourth cusp. small. Cusps 2 (centre) and 3 (right)
Margin of the shell finely crenulate subequal. Cusp 3 with a pointed protu-
7
Ibems, 21 (2), 2003
berance on its outer sideborden Margin II. Morphometric delineation of
of the shell entire or slightly indented species
(Figure 6). External surface smooth at
the apex, with fíat and spaced ribs For all three Patella species, SW and
below. Interior with transparent nacre, SAwas a linear function ofSL (Table II).
often with a green or blue iridescence. However, SH was a curvilinear function
Silveryheadscar. of SL, indicating that as these limpets
Based on shell and radula morpho- increaseinsize, therelativeheightofthe
logical characters reported in the litera- shell increases by a power function.
tureforPatella spp. (Fisher-Piette, 1934, Similar relationship between SH and SL
1935;FiSHER-PiETTEANDGaillard, 1959; has been reported for several Notoacmea
Evans, 1947, 1953; Rolan, 1993; Rolan (Lindberg, 1982) and Patella species
AND Otero-Schmitt, 1996), groups I, II (DAVIES, 1969; Muñoz and Acuña,
and III were unequivocally identified as 1994). Correlation between SH and SL
belonging toP. intermedia, P. idyssiponen- increased after log transformation of
sisandP. indgata,respectively. SH, indicating that an exponential fit
Radula pluricuspid teeth of P. inter- was a better approach than a linear
media from Póvoa de Varzini is similar function.
to P. intermedia type B of Evans (1953), Regressions of SW or SAAvs. SL for
and to those reported by Fisher-Piette P. intermedia, P. idyssiponensis and P.
(1934, figure 2), Fisher-Piette (1935, vulgata specimens yielded slopes not
figure 9, figure 17-2) and Fisher-Piette statistically different (Table II).
ANDGaillard(1959,figure 11 AandB). However, slopes oflog SH vs. SLregres-
Radula pluricuspid teeth of P. sion lines were significantly different
uli/ssiponensis from Póvoa de Varzim is (Table II), indicating that the relative
similar to P. idyssiponensis type A of increaseinheightas theshellgrowswas
Evans (1953), and to those reported by not uniform in these Patella species.
Fisher-Piette (1934, figure 2), Fisher- Slopes increased in the order P. ulyssipo-
Piette (1935-figure 10, figure 17-9) and nensis,P. vulgata,andP. intermedia.
Fisher-Piette and Gaillard (1959 - ForallthreePatella species, RLwas a
figure 14B). linear function of SL and of ''V(SV)
Radula pluricuspid teeth of P. (Table II). Similar relationship between
vulgata from Póvoa de Varzim is similar RLand SLhas been reported for several
to P. vulgata types A and B of Evans Patella species (Brian and Owen, 1952;
(1953), and to those reported by Fisher- Davies, 1969; Sella, 1976). Slopes ofthe
Piette (1934, figure 2), Fisher-Piette regression lines were significantly dif-
(1935, figure 8, figure 17-1) and Fisher- ferent, and increased in the order P.
PietteANDGaillard(1959,figure9A). idyssiponensis, P. intermedia, and P.
P. intermedia, P. idyssiponensis and P. vidgata (TableII).
vuJgata shells from Póvoa de Varzim are In P. intermedia, P. ulyssiponensis and
morphologically similar to those found P. vulgata, radula relative size displayed
in Caliza (northwest Spain) (Rolan, higher variability than shell form.
1993; Rolan and Otero-Schmitt, 1996) Whilst the coefficient of determination
and at Cardigan Bay (Wales, Great for the regressions describing shell form
Britain;EVANS, 1947). (SW vs. SL, log SH vs. SL, and SAP vs.
The foot showed no constant SL) ranged between 0.597 and 0.952,
colouration within each group, and those describing radula relative size (RL
therefore was not used in the Identifica- vs. SL, and RL vs. ^VíSV)) varied
tion. P. intermedia foot was dark, from between 0.472 and 0.767 (Table II).
grey or yellow to black. P. idyssiponensis Whilst the coefficient of variation of the
foot was light, yellow or, most often, means describing shell form (SW/SL,
orange. The colourofP. indgata footwas SH/SL, SAA/SAP, and SAP/SL) varied
very variable, from light yellow to between 5.4 and 22 %, those describing
black. radula relative size (RL/SL, RL/SH,
CabRAI,: Characterization and mulrivariateanalysis oíPatellaspp.in Northwest Portugal
Figures 5, 6. Patella shells from AVer-o-Mar and Agu^adoura beachcs. ^; /.' ulyssipoiiciisis; b: P.
vulgata.
Figuras5, 6. ConchasdeVa.te\\a.provenientesdelasplayasdeA Ver-o-MaryAgugadoura. 5:P. ulyssipo-
nensis; 6: P. vulgata.
Iberus, 21 (2), 2003
Table II. Slopes ofregression lines for morphometric characterization ofthe species, and rheir sta-
tisticalcomparison.
TablaII. Pendientesdelasrectasderegresióndelacaracterización morfométicadelasespecies,ysucom-
paraciónestadística.
Comparisonofslopes
Regression P.intemedia P.ulyssipoilensis P.Vlilgato
(DF=2,604)
Slope r^ Slope r^ Slope r^ F
Shell
SW1/5.SL 0.859 0.952 0.791 0.918 0.821 0.929 3.719^5
LogioSH \is.SL 0.0241 0.744 0.0149 0.597 0.0182 0.754 23.1***
SAPys.SL 0.608 0.859 0.570 0.834 0.627 0.939 2.311^5
Rodula
RL^s.SHL m 1.29 0.472 0.712 0.625 1.40 0.501 8.557***
RL1/5. 2.89 0.485 1.89 0.767 3.15 0.556 5.690***
"^Notsignificantat0.02level
***Significantat0.005level
RL/SAA, and RL/^VíSV)) varied 9.6and tion between species, and the variables
28 % (Table III). Concerning the radula with the highest loadings were RL/SH
relative size, P. iih/ssiponensis was the followed by RL/^'V(SV) (data not
leastvariablespecies (TableIII). shown) The simultaneous use of vari-
ables describing shell form and vari-
III.Discriminationbetweenspecies ables describing radula relative size
(Analysis IV) was better than the use of
In order to determine which vari- one of these groups of variables alone
ables provided the best discrimination (Table IV). However, the improvement
between species, several canonical dis- of Analysis IV over Analysis III was
criminant analyses were carried out limited. Aposterioricorrectidentification
using different techniques and different of P. uli/ssiponensis specimens was
sets of radula and/or shell characteris- unchanged (and very good), but identi-
tics. Results are displayed in Tables IV - fication of P. intermedia and P. viilgata
VIL specimens was lower than 80 7c (Table
Analysis I and II, using only vari- IV). The first canonical variable
ables describingshell form, resulted in a accounted for the great majority of the
very poor discrimination between variation between species, and the
species,with only65 % ofthespecimens second canonical discriminant variable
a posteriori correctly identified to species accounted for only 23.7 % of the vari-
(Table IV). The first canonical variable anee (Table VI). The variables that pro-
accounted for 91 % of the variation vided the highest contrast between the
between species, and the variable with three Pateüa species were RL/SH fol-
the highest loading was SH/SL (data lowed by RL/3V(SV), and then RL/SL
not shown). Analysis III, using only (Table VII). Individuáis were plotted
variables describing radula relative size, along the two canonical variables
was better than analyses I and II, and (Figure 7). There was a considerable
was very satisfactory for P. ulyssiponen- overlap between P intermedia and P.
sis, but resulted in a low a posteriori vulgata specimens, but a reasonable dis-
correct identification of P. intermedia crimination ofP ulyssiponensis individu-
specimens (Table IV). The first canonical áis. Stepwise selection of variables
variableaccounted for 76 % ofthevaria- describing shell form and radula rela-
10