DIFFERENTIATION BETWEEN THE TRACES OF PREDATION OF MURICIDS AND NATICIDS IN SPANISH PLIOCENE CHLAMYS

El análisis estadístico de los orificios perforados en las especies del género ChIamys del Plioceno Inferior del Sur de España aclara algunos aspectos muy interesantes sobre la predacción de los murícidos y natícidoo. El análisis de la función discriminante se ha determinado como un método muy útil para identificar la naturaleza de los orificios, son muy pocos los errores de identificación para los especímenes analizados. En concreto, se ha podido establecer diferencias significativas en las relaciones entre los predadores y las presas. Se ha podido responder afirmativamente a la cuestión de cómo eligen los sitios preferentes para la localización de las perforaciones en el caso de los murídicos. También, para los murícidos la mayoría de las perforaciones observadas (74,1%) se ha concentrado en la región del músculo abductor; en cambio, para los natícid05 no se ha observado este techo (sólo un 36,9%, de perforaciones en la región del músculo abductor). Para 105 natícidos se ha podido dem05trar una asociación entre el lugar de la presa donde se producen las perforaciones y predador, únicamente cuando se toman en cuenta presas de muchas especies y géneros distint05. También existe una estrecha asociación entre el tamaño de la presa y el lugar donde perforan ambos grupos de predadores. No hay una evidencia decisiva que permita correlacionar la localización de las perforaciones para los natícidos, pero probablemente si la haya para los murícidos. La estrategia de selección del lugar de la perforación para 105 murícid05 se ha determinado en función del músculo abductor en 105 monomiari05 ChIamys para facilitar el acceso a las vísceras. En general, 105 métod05 de análisis multivariado y en especial el método de correlaciones canónicas, se han encontrado como de gran ayuda en la interpretación de observaciones cuantitativas.


Introduction
Muricidds and naticids have been important predators oí bivalves, gastropods, cirripedes, ostracods and foraminifers (Livan, 1937;Reyment, 1966) since the early Cretaceous, notwitbstanding tbat tbeir predational history extends mucb furtber back in time (Maddocks, 1988).Inasmucb as tbey leave unequivocal evidence of tbeir predational activity, tbey are valuable ecological indicators.Tbe traces left by these predators in molluscan shells, in particular, are usually identifiable.Tbus, tbe boles drilled by muricids are characterized by baving tbeir external and internal openings of about the same size, witb tbe external diameter only slightly greater wbich, because of tbe straigbt walls of thebole, confer a conical sbape to tbe orifice.Tbe holes drilled by naticids are usually countersunk in sbape, witb the external opening being considerably wider tban the internal opening; tbe walls tend to follow a parabolic course.A detailed appraisal of tbe morphology of gastropod boreholes is to be found in Carriker and Yochelson (1968).
Tbe aims of tbe present investigaiton are directed towards elucidating tbe following points: l.Tbe eventual usefulness of multivariate statistical analysis for s.GUERRERO ALBA, R. A. REYMENT distinguishing between holes drilled by naticids and muricids.2. Tbe association between the size of the prey and the dimensions of tbe borehole: 3. Tbe eventual association between the holes made by the drills in Ch1amys and tbe size of the prey.4. Does tbe size of the borehole in Chlamys show a relationship witb its position?5. ls tbeir a difference in the strategy employed by muricids in tbeir predation on species of CblamyS!

Material and methods
The material studied here was obtained from exposures in the Lower Pliocene of the Río Padrón in the Province de Málaga (Spain) (fig.1).These samples yielded several species of bivalves belonging 10 a number of genera which display abundant evidence of gastropod predation.In addition, species of the bivalve genus Chlamys (are indicated in figs.3,4,5,6,7,8) from various localities in the Lower Pliocene around the city of Málaga were available.These localities are indicated in fig. 1.They are all dated as belonging 10 the zone of GlooorotaJia margaritae.

Univariate or multivariate anaIysis?
We shall begin by addressing ourselves to the question of the utility of multivariate analysis in tbe preseot oonnexion.This can be done oonveniently by oonsidering tbe differences between tbe means for tbe externa!. respectively, internal openings of muricid    muricid boles and 46 naticid boles drilled in CbJamys spp., and 39 muricid boles and 33 naticid perforations in tbe specimens from the Rio Padrón (using only variables L, A, DE and DI).In tbe first case, only 6.5% of tbe naticid boles were incorrectly located witb tbose known to derive from muricids and, on the other hand, even less muricid boles were incorrectly a11ocated, namely, 3.7%.
In general, the linear discriminant function in table 1 for CbJamys provides an efficient means of distinguishing between naticid and muricid predation, as indicated by the highly significant value of the variance ratio FS. 77=23.l3.Tbe use of the corresponding quadratic discriminant function enabled a reduction in tbe levels of incorrect a11ocation to 1.85% for muricids and 6.5% for naticids, whicb is the same value as before.
In the case of the mixed sample of bivalves from tbe Rio Padrón, tbe discriminant function given in table 1 correctly identified allo but 8.2% of tbe naticid boles, and 3.0% of tbose made by muricids.Tbe efficiency of tbis linear discriminant function is indicated by the highly significant value of F m =39.78.Tbe corresponding quadratic discriminant function produced exactly tbe same results as tbe linear function.
As is apparent from table 1, almost a11 of tbe discriminatory effect is due to tbe properties of tbe boreboles, witb little or no input from tbe dimensions of tbe prey.
To answer the question posed at the beginning of tbis section, the multivariate analyses were clearly able lo provide a more nuanced and complete analysis of the data, in addition to giving us a means of allocating new material to one or other of the groups of predators on the basis of measurements on the borehole.

Sizes of predator and prey
Tbe next question we ask is wbether it is possible to establish a relationship between tbe size of tbe prey and that of the predator.Data on the prey are, natura11y, easily enough come by.But how are we to estímated the size of the predator?Tbis can be done, albeit in approximate terms, by making use of the fact that the dimensions of the accessory boring organ (ABO) are closely retlected in the size of tbe borehole.This observation derives from the careful work over more than 30 years of Carriker (1955Carriker ( , 1969Carriker ( , 1981)); also Carriker et al. (1974).Valuable observations on this topic have been published by Wiltse (1980) and the growth-relationship between the ABO and the rest of the dri11 has been used by Reyment et al. (1987) for computing approximate sizes of drills.For oue analyses, we made use of the rather complicated multivariate technique of canonical correlation analysis, in the expanded from, involving redundancy analysis (Coole and Lohnes, 1971), expounded in biological connexions by Pimentel (1979).Tbe analyses based on the two measures on the boreholes and the length, width and thickeness of the she11 of the prey indicated that for both groups of drills, tbe associations witb tbe prey are far from achieving significance.In neitber case does the degree of overlap between either groups (redundancy of one group given information on the otber, i.e., tbe overlap between sets) reach a sufficiently high level for investing the canonical correlation in the entire sample.(N.B. canonical correlations are not representative of ranked fractions of total variation as is the case in principal component analysis.) H we now proceed lo the results obtained for the analysis of the material from the Río Padrón, we find that, once again, there is no significant relationship between the size of the predator and the size of tbe drill for muricids and, in fact, the two sets are no  3 in which it will be seen that there is a alignment in the elements of tbe first principal component and the elements for the vector of morphological structure for three of four variables.6.These results agree closely with those obtained for the muricid holes.The angle between the smallest eigenvectors for both sets of data is only 9°20', thus confirming the similarity in their structures.A possible explanation of the principal component analyses might lie with the fact that the data considered derive from closely related morphologies of the genus ChJamys, noted for the great homogeneity it displays in the dimensions of tbe sbell in its species.

Principal component analysis:
The principal component analyses for the data from tbe Rio Padrón are summarized in tables 7 and 8.The results echo sbarply those found for the ChJamys data in the case of tbe muricid holes.However, sorne aspects are more sbarply outlined.We note that the first vector for muricids is dominated by tbe bole-dimensions, wbereas this vector for naticids is marked by equal participation of all variables.The second vector for muricids is dominated by equal participation of tbe shell-dimensions at tbe expense of the hole-dimensions, wbereas in tbe naticid material, the two shell-dimensions are in a negative relationship with the rougbly equally represented hole-measures.
Further inspection of tbe results in table 7 for the muricid boreholes discloses the interesting detail that each of the four vectors reflects a complete lack of correlation between tbe two sets (tbis is pointed to by the occurrence tbroughout of tbe pairwise loadings of [0.707], [0.707].
Finally, we observe tbat the angle between the two smallest eigenvectors in no more than 2°42, thus indicating almost complete agreement in their structures.

Association between the size of the prey and the location of the driUholes
The same statistical metbodology as used in tbe foregoing section was applied to the present problem.The four variables DE, DI DOU and DOB were weighed against tbe set of measures on tbe sbell of tbe ChJamys, to wit, L, A, G, and the curvature, bere denoted C. The canonical correlation between tbese sets for muricids is bigbly significant, witb X 2 for tbe significance of the rust latent value of tbe canonical analysis reacbing 321 for five degrees of freedom.The significance for the naticids for the same latent values is also bighly significant, being X 2 =239 for five degrees of freedom.The validity of the relationships are strongly supported by tbe exceptionally higb overlaps between sets; tbus, for the muricids, tbe first set overlaps tbe second to tbe extent of 79.92% and tbe second overlaps tbe fll'St by 78.38%.Likewise, we find for tbe maticids, tbat tbe overlap of the second set by first is 66.22% and that by the second set of the first is 70.01%.The respective canonical correlation coefficients are 0.929, for muricid holes, and 0.919 for naticids.
The correlation matrix listed in table 9 shows that tbere is bigh correlation between variable of the two sets and, also, within sets.The naticids (table 9) also display quite high analogous correlations, but these are less marked tban for the muricids.lf we now examine the analytical results summarized in table 10, we observe that the vector for morphological structure for botb sets of boreholes presents the same picture of equal participation of all variables.We also observe tbat the elements of tbe first principal component for botb categories are close indeed.The angle between tbese vectors is 3°38, tbus documenting the agreement in structure.
The results of tbe principal component analyses for botb categories are presented in tables 11 and 12.The good association between the size of the prey and the location of tbe borehole is clearly manifested in both cases.This result would seem to be interpre-

DI .
Eigenvalues: Naticids the present study.If we once again direct attention to the smallest principal component, we note the close agreement in its elements for the both sets of observations.The angle between these vectors is 9°.

Association between the size of the driDhole and its location
Even this particular topie can be best treated in the same statistical manner as in the previous sections.The analysis accounted Cor here was made on the two dimensions oC the borehole balanced against the twoi measures oC its location on tbe shell.
The canonical correlation Cor tbe muricids is not significant, whereas tbat obtained for naticids was determined to be so, with X 2 =22 Cor three degrees oC freedom.The non-significant canonical correlation Cor muricids is reflected in tbe slight degree oC overlap between sets Cor the muricid holes; in the case oC tbe naticids, we have a strongly asymmetric relationship with set one overlapping set two by 20.87% wbereas the value Cor set 2 overlapping set 1 is only 1.58%.The correlation matrix Cor muricids (table 13) indicated tbe between-group correlations to be low; likewise for the naticids (table 13).The canonical vector, morphological structural vector and first principal component listed in table 14 point to certain indicationns that were not shown up by tbe redundancy analysis in that tbe first principal component represents equal covariation in aH Cour variables, a condition tbat is largely reproduced in the morphological structural relationship.A similar result was Cound Cor the naticid holes (table 15), with respect to the agreement in morpbological structure and principal components.However, in their case, the role oC hole-size is very greatly reduced.Thus, tbe evidence seems to point to tbere being an association between the size oC tbe driHbole and its location in naticids, but that this is not so Cor muricids.
The principal component analyses Cor both kinds oC driHholes are summarized in tables 16 and 17.In addition to what has already been mentioned, we point to certain marked differences in the structure oC  the eigenvectors.The first two eigenvectors for the muricids do not differ very greatly with respect to the sizes of their eigenvalues, notably, 56% and 36%, respectively.This indicates that these two principal components represent real relationships in the two sets; thus the first vector seems to say that most variability in the data Hes with bigger holes being located at certain sites.The second vector tells us that an essential part of the sample comprises holes that bear some inverse relationship to the site drilled, presumably indicating that a sizeable proportion of the holes drilled are not directly relatable to the First eigenvalue=49.450%oC tr R Overlap oC second group by first= 1.58% Overlap oC first group by second=20.87%location occupied by the hole in the shell.This conclusion is in harmony with the findings of the redundancy analysis.
In the case of the naticids (table 17), we see that location of the borehole and the size of the borehole are entirely unconnected and that each set has its own principal component, the first dominated by the geometrical location of the borehole, attached to about 49% of the variation, and the second entirely dominated by the dimensions fo the borehole, and connected

Predational strategies of muricids and naticids
As is sbown by the Pope diagrarns in figure 2, the holes drilled by muricids in Chlamys display a pronounced tendency to concentrate in a restricted part of the shell (74.l %), namely, in the rone of insertion of the adduetor muscle; this pattem of distribution of boreholes (only 36.9%) is not seen to tbe same extent in the prey drilled by naticids.On the basis of the present investigation of predation on Chlamys spp., at least, it would seem that muricids manifest a capaeity to select a certain site for initiating drilling.
Sorne species of muricids first drill their prey after wwhieh tbey feed on tbe soft parts directly through the gaping valves.This second behavioural procedure is commonly adopted for feeding on carrion.Suitable prey is provided by monomyarian bivalves, sich a Ch1amys, in which only one muscle requires to be neutralized.The concentration of holes to the location of the adduetor muscle shown in figure 2 is clearly expressed.
Thus, it seerns logical to conclude tbat the drilling behaviour of sorne muricids, in relation to monomyrian bivalves, has sorne kind of non-random component in the sense that it is aimed at weakening the function of the adductor muscle.Naticids are not invested with the same strategy oC attaek since they feed solely by means of the hole drilled.The efficiency of the muricid attack procedure seerns to be relatively high in relaiton to naticids, with respect to the retum oC energy in relation to the effort expended.It is interesting to note that these two veetors are almost orthogonal, thus indicating their completely opposite nature.We may draw a further conclusion, to wit, that the factor of fundamental significance in the predational register of the naticids is the size of the borehole, as the process of feeding takes place through the hole drilled.On the other hand, the behaviour of sorne species of murieids, for example those of Murex, with respect to feeding is different in that the act oC drilling is the means oC obtaining access to the interior of the shell (see, for example, Bames, 1985, p. 379).The size of the hole drilled is therefore not decisive for gaining access to the prey.
The present study has as its aim the elucidation of certain aspects oC gastropod predation, exemplified by the eolleetions at our disposal.It is not meant to be a complete treatise on the general predatory behaviour of the two groups of interest.We have found that some groups of muricids may manifest a preferred interest for drilling a certain site in monomyrian bivalves, but that this behaviour is not obviously ecboed in naticids.Inasmueh as many of the differences between tbe two kinds of boreholes are slight, and undetermined at the univariate level, multivariate statistical procedures have been resorted to; these have been found to yield excellent results and to sharpen greatly the interpretations.
It was thus established that notwithstanding the fact that tbe respective sizes of predator and prey are not usually highly correlated (ef.Reyment, 1966;  Reyment el al., 1978; Ansell, 1960; Thomas, 1976;  Kitchell el al., 1981), one sample did indeed display a high level of such association, notably, naticid holes in material fram the Río Padrón.

Fig
Fig. l.-Sketch map showing the localities samples for the present study.These all lie in the Province of Málaga and are Pliocene in age.
of the variation.Finally, we point to the structure of the smallest principal components: in the case of the muricids, this component is concemed entirely with the dimensions of the borehole, whereas the corresponding component for the naticids is almost entirely a representation of the localization of the borehole.
Pope diagrams displaying the locations of 54 drillholes made by muri¡;ids and 46 by naticids in Chlamys.The dashtd lines denote the probably location of (he adductor muscle.There are 40 muri¡;id holes situattd within the adduetorial Tbe data tbus obtaintd were aoalyzod by standard methods of uoivariate statistics aod by tbe muitivariate statistical methods of linear aod quadflltic discriminant fuDClions, caoonical correlation aoaIysi$, and principal componen\S.

Table l .
-Linear discrimDant function anaIysis ror boles driIIed by muricids and naticids in Cbbunys spp.and the bivalves from the Río Padrón.

Table 2 .
-Correlation matrices, standard deviations and means ror species or Cblamys drilled by muricids and naticids.

Table 3 .
-Correlation matrices, standard deviations and means ror the bivalves from tbe Rio Padrón drilled by muricids and naticids.
more than in contaet with each other.However, in the case of the naticid predation, a completely different situation emerges.The first set overlaps the second to the extent of 35.3% while the second overlaps the first by 38.5%; these values are quite high, and sufficiently so as to indicate that the corresponding canonical correlation is indeed representative of a true association in the vanbles.The relevant details of tbe analysis for the muricids and naticids are listed in table

Table 7 .
-Principal component anaIysís of the correlation matrix for size of prey and drlIIhoIe size for muricids on bivalves from the Río Padrón.

Table 8 .
-Principal component anaIysís of the correlation matrix for size of prey and drlIIhoIe size for naticids on bivalves from the Río Padrón.
table in tbe same manner as before, namely, with respect to the homogeneity in length and width of tbe sbell of the species of Ch/amys encountered in
G .L A e G
L A e G

Table 13 .
-Correlation matrices, standard deviations and means Cor the location and size oC boreholes in ChllUlJys.

Table 16 .
-Principal component analysls of the correlation matrix for sites and sizes of boles driUed by muricids in Cblamys spp.

Table 17 .
-Principal component anaIysis of the correlation matrix for site and size of boles drilled by naticids in Cblamys spp.