taxonomic problems in the genus oliva · 2018-02-07 · phuket marine biological center special...

22
Phuket Marine Biological Center Sp ecial Publication 18(2): 263-284. (1998 ) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA Bernard Tursch Laboratoire de Bio-Ecologie, Faculte des Sciences, Unioersite Libre de Bruxelles, 50 av. RD. Roosevelt, 1170-Brussels, Belgium. ABSTRACT The limitations and the implications of the objective morphospecies approach are briefly reviewed. Attention is called to frequent unwarranted interpretations of data . The mor- phological continuum method is advocated for taxonomic decisions on allopatric samples. Solutions are presented for some problems particular to the genus Oliva, such as the choice of characters and the necessary reduction of the scale of sympatry. THE OLIVA PROBLEM The common and widespread gastropod ge- nus Oliva is quite easy to recognise,because it is homogeneous. Yet, we know nearly noth- ing about the natural hi story of Oliva and over 95 % ofthe literature deals exclusively with nomenclature. Thi s is probably because the formal naming of species is the branch of science that requires the least science. The problem with Oliva is continual shuttling ofthe status of names between those of spe - cies, subspecies,forma and synonym. Within the genus, relatively few species are easy to recognise. Many more 'species' are not easy at all and pose frustrating prob- lems, familiar to tropical malacologists (see Abbott 1991; Tursch 1992). The classical complaint is:"My specimen stands right in the middle between two species". The fre- quency of identification problems certainly means that many (most?) Oliva species have not been properly delimited. I will here confine myself to this prob- lem ofspecies differentation. This is not only an academic exercise: in the tropics, Oliva often constitute a major element of soft sub- stratum communities (see for instance Kohn 1997). The questions of phylogeny, identifi- cation or nomenclature will make sense only after the species problem has first been solved. Difficulties in species delimitation also exist in other groups of molluscs, so some of my problems might also be yours. I will attempt to show that many taxonomic obstacles can be surmounted by using a logi- cal, objective approach . Most Oliva species are notoriously vari - able in some characters, such as size and colour patterns. The combination ofthe two ingredients , homogeneity of the genus and variability ofthe species, is a sure recipe for producingtaxonomic chaos. This has indeed happened: hundreds of nominal taxa are now considered to be species, subspecies , forms or synonyms, apparently at the whim of the authors . Of course many of these names are genuine synonyms. In addition, misidentifications are very common. Inter- estingly, for some species there are very few or no problems. So, from the start , it would seem that one is dealing with a few 'good species' and a large number of 'bad species'. There is no hope ofcleaning up this taxo- nomic mess unless we understand what went wrong in the first place. One main cause of problems is certainly that the defi- nition of simple, reliable taxonomic charac- t er s is very difficult because the shells of many Oliva species are not separated by easy 'yes or no' characters. The shells are devoid of any sculpture and mainly differ by rather subtle variations of shape. These differences are hard to describe in words and at first sight may even appear to be gradual. VARIABILITY OF OLIVA AND VARIABILITY OF TAXONOMISTS The blame for the difficulties in Oliva tax- onomy have invariably been laid upon the

Upload: others

Post on 22-Apr-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

Phuket Marine Biological Center Sp ecial Publication 18(2): 263-284. (1998 ) 263

TAXONOMIC PROBLEMS IN THE GENUS OLIVA

Bernard TurschLaboratoire de Bio -Ecologie, Faculte des Sciences, Unioersite Libre de Bruxelles,

50 av. RD. Roosevelt, 1170-Brussels, Belgium.

ABSTRACTThe limitations and the implications of the objective morphospecies approach are brieflyreviewed. Attention is called to frequent unwarranted interpretations of data. The mor­phological continuum method is advocated for taxonomic decisions on allopatric samples.Solutions are presented for some problems particular to the genus Oliva , such as the choiceof characters and the necessary reduction of the scale of sympatry.

THE OLIVA PROBLEMThe common and widespread gastropod ge­nus Oliva is quite easy to recognise, becauseit is homogeneous. Yet, we know nearly noth­ing about the natural history of Oliva andover 95 % ofthe literature deals exclusivelywith nomenclature.This is probably becausethe formal naming of species is the branchof science that requires the least science.Theproblem with Oliva is continual shuttlingofthe status of names between those of spe ­cies, subspecies, forma and synonym.

Within the genus, relatively few speciesare easy to recognise. Many more 'species'are not easy at all and pose frustrating prob­lems, familiar to tropical malacologists (seeAbbott 1991; Tursch 1992). The classicalcomplaint is : "My specimen stands right inthe middle between two species". The fre­quency of identification problems certainlymeans that many (most?) Oliva species havenot been properly delimited.

I will here confine myself to this prob­lem of species differentation. This is not onlyan academic exercise: in the tropics, Olivaoften constitute a major element of soft sub­stratum communities (see for instance Kohn1997). The questions of phylogeny, identifi­cation or nomenclature will make sense onlyafter the species problem has first beensolved. Difficulties in species delimitationalso exist in other groups of molluscs, sosome of my problems might also be yours. Iwill attempt to show that many taxonomicobstacles can be surmounted by using a logi-

cal, objective approach.Most Oliva species are notoriously vari­

able in some characters, such as size andcolour patterns. The combination ofthe twoingredients, homogeneity of the genus andvariability ofthe species, is a sure recipe forproducing taxonomic chaos. This has indeedhappened: hundreds of nominal taxa arenow considered to be species, subspecies,forms or synonyms, apparently at the whimof the authors. Of course many of thesenames are genuine synonyms. In addition,misidentifications are very common. Inter­estingly, for some species there are very fewor no problems. So, from the start, it wouldseem that one is dealing with a few 'goodspecies' and a large number of 'bad species'.

There is no hope of cleaning up this taxo­nomic mess unless we understand whatwent wrong in the first place. One maincause of problems is certainly that the defi­nition of simple, reliable taxonomic charac­ters is very difficult because the shells ofmany Oliva species are not separated byeasy 'yes or no' characters. The shells aredevoid of any sculpture and mainly differby rather subtle variations of shape. Thesedifferences are hard to describe in words andat first sight may even appear to be gradual.

VARIABILITY OF OLIVA ANDVARIABILITY OF TAXONOMISTS

The blame for the difficulties in Oliva tax­onomy have invariably been laid upon the

Page 2: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

264 Tropical Marine Mollusc Programme (TMMPj

'extreme variability' of the shells (which isa biological problem). But could it be thatpart of the problem is due to the variabilityof the taxonomists?

Just for fun , this hypothesis was te stedat a meeting, many years ago, by asking 29experienced shell collectors to say how manyspecies there were in the same sample of 12Oli va shells. The fact that the panel con­sisted of hobbyists matters little, as mostOliva taxa have been described by non-pro­fessionals . The distribution of answers wasvery irregular, with a large standard devia­tion. The correct answer (which I learnedmyself many years later ) is different fromthe mean, the median and the mode (illus­tr ating, by the way, that taxonomic mattersshould not be settled by a majority vote ).Thecoefficient of variation was over 50 %, waymore than what could be measured on theactual specimens. So, in this particular case,malacologists were more variable than theOliva .

This variability reflects methodologicalproblem s and there are many. The first iscommon to all of malacology: many of thenames are due to the descriptive excess ofsome authors. It is true that taxonomists aresupposed to describe and classify variation,but they are certainly not supposed anymoreto name all variants in terms of arbitraryspecies and subspecies. This attitude is, ofcourse, re lated to an antiquated typo logicalconcept of species. This concept unfortu­nately still thrives today, albeit often mas­querading under the vocab ulary of evolu­tionary biology.

The second methodological problem is theuse of a vague vocabulary to describe veryvague taxonomic 'character s'. Here is, forinstance, the description of an Oliva species:"Shell variable in shape, us ually fusiform,with rounded sides and low spire; colorranges from white, yellow, orange, cream orwhite, to brown and black, overlaid withdark brown or black zig-zags, triangles,blotches, and/or dots; aperture dark brown;columella white with dist inct , although

poorly-developed plications." The text (au­thor deliberately not cited) calls for severalremarks:

* The description hardly contains anyone character that is well -defined and use­ful. All Oliva are "usually fusiform" and have"rounded sides" to some extent. The colourpattern is so general that it could also de­scribe my cat. The only clear indications "ap­erture dark brown; columella white" are sim­ply misleading: there are many exceptions.

* This text dates from the mid-eightiesbut could also be 150 years old. It gives onethe feeling that nothing has happened in bi­ology dur ing the last two centuries.

* This is not the language of the scien­tist, but more that of the art criti c. Thi s canonly lead to a 'semi-aesthetical' classifica­tion. I have nothing against aesthetics (wellon the contrary) but such a classification isnot refutable as scientific statements shouldbe.

So, in practice, it is im possible to recog­nise most Oli va species from their wordeddescription without the help of an illustra­tion or of ty pe material.

The third methodological problem con­cerns the traditional choice of characters.The paramount characters used in descrip­tions of Oliva species are: absolute size , theshape of the body whorl, the relative heightof the spire, the presence of a callus on thespire, the number and the aspect of columel­lar plica tions and the colour pattern of thebody whorl. The choice of these 'classicalcharacters' is most unfortunate: it so hap­pens that they are precisely the most vari­able she ll attributes in the genus Oliva.

* Size is highly variable. Small and largethick-lipped specimens can be found withinthe same population (see Fig . 1). Mean sizeof popu lations is generally not centred onlarge specimens (see Tursch et al. 1995) andthere is no way of determining from the shellif a specimen is 'adult' . For writers on Oliva ,this word is often used with a rather vaguemeaning.

* The shape of body whorl and the rela-

Page 3: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998 ) 265

Figure 1. Variability in size (sse text p. 2). Example: two thick-lipped specimens of O. miniacea(Roding, 1798) from Kwajalein Atoll, Marshall Is. Scale bar: 10 mm.Figure 2. Variability in relative spire height and shape of body whorl (see text this page). Example:three thick-lipped specimens of 0. flammulata Lamarck, 1811 from Luanda, Angola.

tive height of spire are very variable in manyspecies (see Fig. 2). The relative height ofthe spire is possibly the most variable meas­urement in Oliva (see also Figs. 4, 5, 6).

* The presence of a callus on the spire isnot always a reliable character. In some spe­cies, one can find all intergrades betweenfully callused and completely open spires,sometimes even in the same population (seeFig. 3).

* Columellar plications can vary verymuch in importance, even amongst speci-

Figure 3. Variability in spire callosity (see textthis page). Example: two thick-lipped specimensofo. irisans Lamarck, 1811 from the same popu­lation, Pulau Hantu, Singapore.

mens of similar size within the same popu­lation (see Fig. 4). For some species they dohave a rather characteristic aspect, but inmost cases, the plications are difficult tocount because they get progressively smallerand indistinct. In general, small specimenshave sharper, more distinct columellarplaits.

* Colour and patterns on the body whorlmay in most species be extremely variable(see Fig. 5). This aspect will be discussedmore in detail later.

Except for the size, these features havealways been reported in subjective terms. Sothe variability of these 'classical characters'is further amplified, because their descrip­tion is often as variable as the character it­self. For example, for a long time there wasonly one known specimen of Oliva esiodinaDuclos. The spire of this specimen was de­scribed by three authors, each in a very dif­ferent way (see Tursch 1992).

METHODOLOGYThe species conceptIfmuch of the 'Oliva problem' derives frominadequate methodology (or no methodologyat all) , the first task is to define an opera­tional system for detecting the limits of spe­cies (identification, a different question, can

Page 4: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

266 Tropical Marine Mollusc Programme (TMMP)

Figure 4.Variability in columellar plications (seetext p. 3). Example: tw o thick-lipped specimensof O. miniaceaflammeacolorPetuch and Sargent,1986 from the same lot , 20 fms off Madras, India.

only come later).The more complex the prob­lem, the more we have to be strict about prin­ciples. This is certainly the case for the ge­nus Oliva.

At this stage, two points are already clear.First, most species have been very poorlydefined and, in any case, the species conceptof many early authors was very differentfrom ours. So we'd better start from scratch,without trusting the Wisdom of the Eldersand admitting any species a priori. Second,the best way to prevent Oliva taxonomy frombeing largely a matter of personal opinionis certainly to use a quantitative approach.

At the species level, a logical classifica­tion is a two-step procedure. The first step

consists in grouping similar animals intophena (sets of animals that resemble oneanother). The objective existence of thesephena is only a working hypothesis at thestart. The validity of this hypothesis mustthen be tested by checking if these phenacan or can not be objectively separated fromothers. Phena that cannot be separatedmust, of course, be combined into largerunits.The second step (ranking) consi sts indetermining the taxonomic rank (species,subspecies, form ) of the groups that havebeen separated (or united) in the first step.Only then does naming make any sense.

Before delimiting species, let us first .agree on what we shall consider a species.For practical reasons we are using the mor­phospecies. This word has different mean­ings to different authors. It is used here inthe meaning "objective morphospecies"(which can be demonstrated), in oppositionto what we might call the "psychospecies"(which is an act of faith). The "objectivemorphospecies" is not a new concept of thespecies. It is only an indirect approach to thebiological species (which is not operationalin practice: has any species of marine mol­lusc ever been defined by direct proof of re­productive barriers?).

The morphospecies approach is basedupon the demonstration of gaps in the dis­tribution of phenetic characters. One sim­ply determines the ranges of variability oftwo or several phena; if the ranges do not

Figure 5. Colour patterns. Some species are extremely variable (see text p. 3). Example: six thick­lipped specimens of 0. bulbosa (Roding, 1798) from the same lot, Tulear, Madagascar. Scale bar: 10mm.

Page 5: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

Phuket Marine Biologi cal Center Sp ecial Publication 18(2): 263-284. (1998) 267

Figure 6. Colour patterns. Some populations are very homoge­neous (see text p. 3). Example: three thick-lipped specimens of0. picta Reeve, 1850 from the same lot, 35 m, Hansa Bay, PapuaNew Guinea. This is a case of cryptic colouration: specimensmatch the colour of the sediment (dark sandy mud with smalldebris). Scale bar: 10 mm.

overlap, the phena are separated by gaps.The existence of such morphological gapscan demonstrate the absence of hybrids, butonly under certain conditions (see Tursch1997 ). It will be seen later that these condi­tions are not always met. The morphospeciesapproach has its limitations. It cannot de­tect true sibling species (these can be evi­denced mainly by molecular or behaviouralstudies) and is thus necessarily minimal­istic.

So, instead of lamenting about the vari­ability of Oliva , let us now take full advan­tage of it and use it as an efficient tool. Vari­ability is indeed the worst enemy of thetypologist but the best friend of the evolu­tionist.

Detection ofgap sBefore discussing the biological meaning ofgaps, one should first make sure that thesegaps reflect reality. Evidencing gaps in met-

Figure 7. Colour patterns: convergence by crypsis (see text p. 11). Example:O. caerulea (Roding, 1798) (left) and O. concinna Marrat, 1870 (righ t) .Populations living on white coral sand with debris, Hansa Bay, Papua NewGuinea. Scale bars: 10 mm.

Page 6: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

268 Tropical Marine Mollusc Programme (TMMP)

Figure 8. Colour patterns: convergence by crypsis (see text p. 11). Example: O.elegans Lamarck, 1811 (left) and 0. reticulata (Roding, 1798) (right) . Populationsliving on dirty coral sand and rubble, Mandi, Hansa Bay, Papua New Guinea.Scale bars: 10 mm .

ric characters looks very sim ple but raises alot of problems, both in theory and in prac­tice.

One has to keep in mind that the size ofthe gaps does not necessarily have to belarge. In the genus Oliva , most of the ob­served interspecific gaps are smaller ormuch smaller than the intraspecific variabil­ity ranges. This is what gives the impres­sion that some species form a continuum. Italso entails the necessity of very accurateobservations.

Only full separations do constitute gap s.A difference in mean values, even when sta­tistically significant, is no evidence for theabsence of hybrids (it can, at best, mean thathybrids are rare).

Only features (or combination offeatures)that yield separations do constitute taxo­nomic characters.The others, no matter howinteresting, are just features.Note that a fea­ture that is not a character at the specieslevel might be an excellent character at thegenus level.

By the way, it is often believed that themost interesting taxonomic characters arethose separating all the objects being stud­ied . Such an analysis actually has the sameinform ation content as one that separatesnothing. The highest information content isfound when half the objects are separated

AJlFigure 9. Analysis on one variable (one dimen­sion: frequencies hi stogram). See text this page.

and the other half are not. This does makesense because classifications rest upon bothseparations and combinations.

In biometric studies, each specimen isrepresented by one point in a space havingas many axes as there are characters to beconsidered (t h e attribute hyperspace ).Phena are thus re presented by clusters(clouds') of points. The difficulty in detect­ing gaps between the phena varies with thenumber of dimensions of the attribute space(this is the number of variables that we wantto observe at the same time). One shouldnotice that groups separated on one, two,three or x dimensions are by that very factalso separated in the attribute hyper space.

Separations on one variable (histogramsof frequencies, see Fig. 9) are very familiarand pose no particular problem. Because thegenus Oliva is very homogeneous, many spe­cies are packed into a small range of totalvariation. So it is fully expected that good

Page 7: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 269

Figure 10.Analysis on two variables x and y (twodimensions: scatter diagram ) can separate enti­ties that are not separable by x or y taken sepa­rately. See text this page.

unidimensional separations will be observedonly in lucky cases.

Separation on two variables (scat ter dia­grams) are also very familiar. One will no­tice that the combination of two variables xandy can separate entities that are not sepa­rable by x or y taken separately (see Fig. 10).In the same way, the combination of threevariables x, y and z can separate entities thatare not separated by combinations of the

variables two by two (see Fig. 11).Should we then extrapolate these obser­

vations and simultaneously use as many di­mensions as possible? Use the whole at­tribute hyperspace? There are several rea­sons for moderating our taxonomic spatialambitions.

Taxonomists are essentially visual ani­mals and, by increasing the number of di­mensions, we are going to meet severe prob­lems of representation. There is no problemwith one or two dimensions (we are used tothese).With just three dimensions, it alreadytakes some fiddling to orient the image inorder to show a convincing separation (as inFig. 11). With more than three axes, we can­not see anything at all :we are in hyperspace.Of course, we can use a whole arsenal ofma­thematical tricks such as factor analysis,which reduces the number of axes by mak­ing linear combinations of variables. But theresulting image can be severely distortedand much information can be lost. The prob­lem is specially serious if we include vari­ables that are not proven taxonomic char­acters: they can easily generate enough 'ran­dom noise' to obscure real separations. This

zy

z

x

x~~.

Figure 11. Analysis on three variables x, y and z (three dimensions diagram) can separate enti tiesthat are not separated by combinations of the variables two by two. See text this page.

z

Page 8: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

270 Tropical Marine Mollusc Programme (TMMP)

does not mean that we should not use factoranalysis. On the contrary, it is most useful,mainly for detecting correlations betweenvariables and thus for eliminating taxo­nomically useless features.

In the attribute hyperspace, one can gethorribly lost. In addition to problems of rep­resentation, there is a problem with the den­sity of points. Fifty shells in a two-dimen­sional representation look like a nice littlecrowd and make us feel quite confident inour analysis . The same fifty specimens in a10-dimensional space can become so scat­tered as to make interpretations qu ite un­safe. To get the same density of points (andthe same feeling of confidence) we had be­fore , we would have to measure millions ofspecimens. In my opinion, bidimensionalscatter diagrams are the best practical com­promise. These graphs are very heuristic,clearly show the 'special covariance sets' inwhich the taxonomist is interested (Gould1984) and large quantities of scatter dia­grams, us ing different combinations ofvari­ables, can be very rapidly computer-pro­duced.

The examples of Figs. 9, 10 and 11 aretoo simple and hide another, very seriousproblem. In practice, we are not dealing withnice surfaces or volumes but with scatteredexperimental points.We draw these reassur­ing envelopes only after we have decidedwhich points go together and which do not.This is the important problem of clustering.There are many different mathematical clus­tering algorithms (see Sneath & SokaI1973),mainly aiming at obtaining 'taxonomic dis­tances' (a notion not used here). These dif­ferent algorithms often give different solu­tions in complex cases. So, what does simplecommon sense say?

Nobody will hesitate much in admittingthe existence of a gap in the scatter diagramof Fig. 12A. Let us now bring the two groupscloser to each other, as in Fig. 12B. The sepa­ration is not so reliable anymore. It couldwell be meaningful, but there is a chancethat any additional specimen could appear

right in the middle of the 'gap'. Our confi­dence in the reality ofthe gap would greatlyincrease if it persists when the number ofspecimens is augmented.At the species level,our confidence would also increase if thesamples contained specimens from manydifferent localities: more of the variabilitywould then be included in the analysis.

Let us then bring the groups even closer,as in Fig. 12C (it has earlier been arguedthat gaps can be very small).The two groupsare still perfectly separ ated, as statisticalte sts would show. We could even fit a verygood discriminating function somewherealong the dotted line. Such separations, fre­quently presented as 'proofs' , are really mostunconvincing, for two reasons. First, the un­certainty about the position of additionalpoints is even greater than in the previouscase. Second, the very same distribution ofpoints could also be perfectly separated inanother way, as in Fig. 12D (with equallyconvincing statistical backing and the bless­ing offactorial discriminant analysis).

There is one easy way out of the 'verysmall gap problem'. In Fig. 12E, the differ­ent symbols (full and open circles) indicatethat we have additional information, this isany kind offeature that does not automati­cally derive from the values of x or y, Forinstance, ifx andy are shell measurements,let us say that the full and open circles rep­resent different colours of eggs (it could bedifferent localities, etc .) Now, the separationpresented in Fig. 12E becomes highly cred­ible because it is supported by facts thatwere not even used in the separation proce­dure. If one makes a statistical test, the col­our of eggs is highly correlated to variablex. This is not a mathematical, but a biologi­cal correlation: exactly the kind of informa­tion the taxonomist is looking for.

No sacred book says that a discriminat­ing function should be linear. If it is not, asin the case of Fig. 12F, the use of additionalinformation is, here again, the only practi­cal way out.

Evidencing the absence of gaps, ie show-

Page 9: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

Ph uket Marine Biological Center Special Publication 18(2): 263-284. (1998) 271

y

f!jy

OK OK?

f!jf).• •

f).• •A )( B )(

y

not OK

y

not OK

.........~.....

~ .....

y

c

E

x

OK

~iOi/~Additional evidence!

x

y

o

F

x

x

Figure 12. Problem s of clustering. See text on opposite page.

Page 10: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

272 Tropical Marine Mollusc Programme (TMMP)

x

CHOICE OF PHENETICCHARACTERS

Figure 13. The problem of overlap. See text thispage.

shell characters will be considered here,mainly because many recent taxa are knownfrom their shell alone. Shell characters oftaxonomic value in the Olividae are of twokinds: colour patterns and shell morphom­etry.

Colour patternsThese have been the source of so many er­rors that they deserve special consideration.There are two main problems with colourcharacters : complexity and variability.

Complexity: The patterns of many Olivashells are often very intricate, defying accu­rate verbal description or even simple cod­ing . Furthermore, all Oliva specimens be­ing different, what should be described is notthe pattern but the 'style' (the program thatleads to the pattern).This is not an easy taskbecause different sets ofparameters can leadto nearly identical patterns. The nature ofwhat we see is also complex : it is the addi­tive effect of different patterns occurring indifferent crystal layers. The final aspect de­pends on the intensity of colours, the thick­ness and the transparency of the layers.

Variability: The range of variation is of­ten very great, but always within limits.Even the most variable species of Oliva hasa large, but not unlimited repertoire of col­our patterns. Any pattern of the repertoirecan be expressed or not (for reasons gener­ally unknown to us ). So colours and patternsoften appear to be 'optional' features. Thereare some regularities in the variability ofcolours and colour patterns. Colours andpatterns are most probably not even per­ceived by the Oliva themselves. But theycould be important in the relation betweenOliva and their predators.

All intermediate situations are met be­tween two extreme "colour strategies":crypsis and polymorphism. In many species,all the members of a population are quitehomogeneous in aspect (see Fig. 6). The col­ours of the shell (and of the soft parts) thenoften match the colour of the sediment (seeVan Osselaer et al. 1993). Most biologists

Overlap OKy

CharactersThe phenetic characters we need for themorphospecies approach could be found ei­ther in anatomy or in shell characters. Only

ing an overlap in the distribution of charac­ters, is as important as evidencing presenceof gaps. Demonstrating overlaps is mucheasier than demonstrating separations. Theoverlap illustrated in Fig. 13 is quite con­vincing, although resting upon a very smallsample. This overlap will indeed persist, nomatter how many additional specimens areincluded in the analysis (only full separa­tions constitute gaps, as previously shown).

The extrapolation of separations ob­served on limited samples always includesa risk oferror, so the delimitation ofmorpho­species is, to some degree, probabilistic. Butwe certainly can increase our chances of be­ing right. In short:

* Separations with small gaps requirelarge samples.

* Separations with very small gaps re­quire 'additional information'.

* Overlaps can always be trusted, evenbetween small samples.

Page 11: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 273

then speak of cryptic coloration, supposedto protect from predators. Note that, in suchcases, the general colour describes thesubstrate more than the animal. It followsthat when two different species 'imitat e' thesame substrate, they also necessarily 'imi­tate' each other (see Figs. 7 and 8). So theynot only fool predators, but they also fool tax­onomists into believing they are the samespecies. Note that, in such cases, the ven­tral faces ofthe shells (normally not seen bypredators) do often differ substantially.

In many other species, every specimen ofa population is different in colour (see Fig.5). Many biologists then speak of polymor­phic coloration, supposed to confuse the'search image' of predators (the scepticalreader might notice that we always have agood story ready, whatever the situation). Inthis case again, the Oliva do not only foolpredators; they also fool taxonomists into be­lieving that they deal with different species.The same confusion might occur in the caseof the 'crypt ic species' that can live in verydifferent sediments, each time adopting adifferent aspect.

So, in many cases, colours and generalcolor patterns are unreliable specific char­acters because they are too sensitive to en­vironmental influences.The nature oftheseinteractions is still far from being clear. Olivaoliva (L., 1758 ) living on blade sand beachesare very largely blackish. We have shownexperimentally (Van Osselaer, Bouillon &Tursch et al. 1993) that black 0. oliva do notchoose between two different sediments;they will go indiscriminately into black orwhite sand. Many Oliva specimens in col­lections have abruptly switched from onepattern to another. Such changes have beenexperimentally induced (Tursch et al. 1995).When black O. oliva taken from black sandwere kept on white sand, nearly all sharplyswitched from black to white color (one un­solved question amongst many: how do thecolour-blind Oliva 'know' the colour of thesubstratum?).

It is tempting to speculate that patterns

are biologically more important than colours .The in situ colour of an Oliva living below15 m depth is certainly not what we see inour drawers (the reds and yellows are gone ).And, in general, most of the many potentialOliva predators have no colour vision butmany can distinguish patterns. Patternsmay also have adaptive values: strong con­trasting colour zones (disruptive patterns)can visually break up the shell outline; oth­ers (aposematic patterns) can be warningsignals protecting toxic animals. One shouldalso keep in mind that some patterns canbe cryptic when seen at a distance but con­spicuous at close range.

I do not downplay the taxonomic impor­tance of colour patterns. They often consti­tute our first, immediate clue to identifica­tion. But identification only makes sensewhen we already have a good classification.In any case, optional and variable charac­ters such as colour patterns would lead to aclassification based upon polythetic taxa. Insuch taxa, organisms that have the greatestnumber of shared characters are placed to­gether, but no single character is either nec­essary or sufficient for group membership(see Sneath & Sokal 1973). Such a classifi­cation might well be sensible but would befundamentally different in nature from allwe are used to (it would for instance invali­date clustering by synapomorphy, see Pan­chen 1992 ). In contrast, biometry leads tothe familiar monothetic taxa in which thepossession of a unique set of features is bothsufficient and necessary for membership.

Shell measurementsWe are thus reduced to search for shell char­acters in morphometry. It has already beensaid that Oliva shells are devoid of meristiccharacters (these are discrete variables thatone can count, such as a number of spines,for instance). So we are restricted to metriccharacters (these are continuous variablesthat one can measure, for instance the di­ameter of the shell).

Many shell measurements specially

Page 12: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

274 Tropical Marine Mollusc Programme (TMMPJ

..---D-'

\\

BUT

Teleoconch,apical view

Teleoconch, latera l view

MPRO.----.

Protoconch,apical view

Protoconch,lateral view

Figure 14. Sketch of some shell measurements. See text this page.

suited for Oliva studies have been definedin this laboratory (Tur sch & Germain 1985 ,1986, 1987;Tursch & Van Osselaer 1987;VanOsselaer & Tursch 1988, 1993; Tursch &Bouillon 1993). Some are shown in Fig . 14.This figure is only a rough sketch: the meas­urements are accurately defined and manyrequire precise orientation of the shell. Allhave been tested for reproducibility, preci­sion and taxonomic potential. These featurescan also be measured on fossil material.

In theory there is an infinity of possiblemeasurements, but there are practicallimi­tations. The problem is not with the datathemselves, as modern microcomputers canrapidly handle very large databases. Theobstacle is in measurements: if it takes halfa day to measure one specimen then the lifeexpectancy of taxonomists might constitutea real limitation.

Measuring shells and handling morpho-

metric calculations takes much time andeffort and is not adapted to field work. Oursecret hope was that, once species have beenproperly defined on morphometric criteria,we would find some 'easy character' to tellthem apart at first glance.

The use ofshell measurementsShell measurements fall into two differentcategories: intensive variables (that do notdepend on size) and extensive variables (thatare depend ant on size) .

Protoconch measurements (obtainedfrom camera lucida drawings made undermagnification) are always considered to beindependent ofthe size of the teleoconch. Butis this really true? The fact that the proto­conch does not grow further after its com­pletion does not ensure that it's character­istics are not correlated with shell size . Onecould, for instance, imagine that larvae with

Page 13: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

Phuket Marine Biological Center Special Pu blication 18(2): 263-284. (1998) 275

NW

4.2

O. mantichora

~'.\. '.:..:::.

O. amethystina

O. parkinsonia

3.7

3.2

LPRO

Figure 16.After objective delimitation of the spe­cies by biometry, one can generally find clues forquick identification (see text this page). Exam­ple: Oliva am ethystina (Roding, 1798) (a) is re ad­ily separated from the closely related O. man­tichora Duclos, 1835 (b) by the colour pattern inthe suprafasciolar zone.

Fig. 15.An example of separation by direct use ofintensive variables (see text this pa ge). Oli vaamethy st ina (Reding, 1798), O. mantichoraDuclos, 1835 , O. bu elowi Sowerby, 1889, O.parkinsoni Prior, 1975. Samples (each n>15) se­lected over whole distribution ranges. Scatter dia­gram of pr otoconch measurements NW vs. LPRO(see Fig. 6). Minimum convex polygon s.

a larger protoconch would turn into individu­als with a larger adult shell. Massive experi­mental evidence established that this is notthe case in Oliva. Protoconch measurementsare intensive (size independent) variables.They are also independant of sex . So, we canuse protoconch measurements directly assuch.They do have a great power of discrimi­nation, as shown in the example of Fig. 15.

This example gives a good opportunityto illustrate that highly accurate microsco pemeasurements are not always required inOliva identification . Oliva amethystina(Roding, 1798) and O. mantichora Du clos,1835 that are well separated in the graph ofFig. 15 (and in many other graphs) were forcenturies confused under the name O. annu­lata Gmelin, a nomen dubium (Tursch et al.1986). Their shape and colour patterns areindeed quite similar and (of course) veryvariable. Now that their separate specific

0.45 0.55 0.65 status has bee n established, the two speciescan be dist inguished at first glance by theircolour pattern in the suprafasciolar region(see Fig. 16). O. am ethystina has squarish,diffuse purplish blots of variable intensity(generally in small number and sometimesabsent); O. mantichora always has thick,sharply defined, curved lines (dark purpleto brown, generally numerous). Note that weare not using general colour pattern but adetail of ornamentation. It concerns a verysmall portion ofthe shell, located on the ven­tral side (not normally exposed) and seemsunlikely to have any adaptive function.

The protoconch of Oliva shells is oftenbroken and sometimes difficult to observe(for instance if it is hidden und er a callus).So we also nee d to use teleoconch measure­ments. In contrast to protoconch measure­ments, all linear measurements of the te leo­conch do obviously vary with the size of theshel l. They are extensive variables . Massiveexperience has shown that if one uses thesemeasurements as such in scatt er diagramsor in factor analysis, one will mainly sepa­rate big shells from small shells .And we al­ready know that the shell size of Oliva isextremely variable (and lea ds to many prob­lems).

There is a nice way to dodge this prob­lem of size . If in stead of using raw linear

Page 14: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

276 Tropical Marine Mollusc Programme (TMMP)

measurements one use ratios ofthese meas­urements, one now has descriptors of shape,in theory more independent of size . Theseratios not only make more taxonomicalsense; they also have a much greater dis­criminating power. This might appear to bea rather strong statement, so it requires ademonstration. Let us take the example ofthe linear teleoconch measurements H, Dand R (data in Tab. 1), taken on two groupsof shells (Group 1 and Group 2, from differ­ent localities).The classical approach wouldconsist in examining the bivariate diagrams(here: regression lines) of the pairs of meas­urements Hand D (Fig. 17A), Hand R (Fig.17B) and D and R (Fig . 17C). From thesegraphs, it is far from evident that there isany difference between Groups 1 and 2. Incontrast, their separation is obvious in thescatter diagram of Fig. 17D which uses thevery same measurements, expressed as theratios HIR and HID (shape factors).

INTERPRETATION OF GAPSIt is one thing to observe gaps (and we havenow rather sensitive tools to do this) but itis another to interpret these gaps (or theirabsence) in biological terms.

Biological interpretation of 'no gap 'As long as we are unable to give any objec­tive evidence for the existence of a gap sepa­rating two samples then, in all circum­stances, these samples must be consideredto belong to the same morphospecies. Spe­cies are not a product of our minds, they areobjective, natural entities. We are not sup­posed to invent a natural order; we are sup­posed to demonstrate its existence. Weshould therefore admit the existence ofsepa­rate species only when there is no way ofavoiding it .

Let us however keep in mind that taxo­nomic decisions are always temporary be­cause there is an infinity of possible charac­ters and all possible characters have notbeen tested. New facts may (and will ) bringchanges in our interpretations. Like all other

Table 1. The use of shape factors (see text this

page). Numerical data for example of Fig. 9.

H D R HID HIRGroup 110.00 6.00 5.50 1.67 1.8250.00 29.80 26.00 1.68 1.9235.00 21.00 18.20 1.67 1.9222.00 13.00 12.50 1.69 1.7612.00 7.00 6.50 1.71 1.8531.00 19.00 17.30 1.63 1.79

Group 28.00 5.30 4.00 1.51 2.00

60.00 40.00 29.00 1.50 2.0740.00 26.00 19.50 1.54 2.0537.00 24.00 18.30 1.54 2.0215.00 10.00 7.40 1.50 2.03

sciences, taxonomy is dynamic and mustconstantly adapt to new evidence. So, inshort: "no proof of gap, no morphospecies".This logical restriction results, of course, ina sharp diminution in the number of admit­ted species. Of course, if some specimensseem 'different ' or 'interesting' enough wecan give them any name below the specieslevel, just to keep in mind that further studymight be necessary.

Anybody who applies the method advo­cated here will be called a 'lurnper' (oftenwith some commiseration by collectors who'know' their shells). Please note that 'split­ters' and 'lumpers ' can exist only when taxo­nomic decisions are a matter of personalopinion or are based on inadequate data.This can certainly be the case at the supra­specific level: one can disagree on where to'cut the branches' of the supposed phylo­genetic tree (your subfamily could becomemy genus). But the species is (or at leastshould be) the most objective of all taxonomiccategories. At the species level, we shouldnot even have the choice between 'splitting'and 'lumping'. These two attitudes can, atbest, be provisional strategies for handlingunsolved cases.

Biological interpretation ofgapsDifficulties in the taxonomic interpretationof morphological gaps may stem from a va­riety of causes. In many animal species

Page 15: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

Phuket Marine Biological Center Special Pu blication 18(2): 263-284. (1998) 277

40

20

o

A

o•

H

20

10

R

o

B

o •

H

20

R

20 40

o •60

1.70

1.60

HID

10

o

20

c

40

o

1.50

1.8

o

1.9

H/R

2.0

Figure 17. Analysis of intensive variables. A, B, C (see text p. 14): Comparison of scatter diagrams ofdirect measurements. D: scatter diagram of reduced measurements (shape factors) . Data from Tab. 1.

adults and juveniles, males and females andeven different populations of a same specieshave very different morphological character­istics. So the customary 'common sense'statement "if two samples are demonstra­bly different they must be of two separatespecies" is totally unjustifiable on theoreti­cal grounds. Such a 'conclusion' is really onlya gue ss (unless the observed differences are

very important, let us say beyond the varia­tion range within a genus). Of course, ifsub­sequently demonstrated by objective re­search, some guesses can turn out to be right.Many do not.

Morphological gaps do clearly demon­strate the existence of separate species onlyif four conditions (at least) are met (seeTursch 1997 ).

Page 16: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

278 Tropical Marine Mollusc Programme (TMMP)

••

L (mm)

O. buetowi

••

14 D (mm)

..'

~/ ... / ..12

. ........ .~.ji". ;········D ·

: .( '".

10 Q. ,~~~......~.......

•••...1 •8 ,' ., / . , .

L (mm)

.,'". > , @

.....,;'".

(mm)

70

10

90

30

50

110

5 25 45 65 85 105 14 18 22

Figure 18. Th e growth of many Oliva species isapproximately isometric (see text this pa ge). Ex­ample: O. porphyria (Linnaeus, 1758), samplefrom West Panama .

Figure 19. The growth of some Oliva species isnon-isometric (see text this page). Example: O.buelowi Sowerby, 1889 , sa mple from Hansa Bay,Papua New Guinea.

(1) The samples to be compared must bestrictly sympatr ic (living in the same area).Morphological gaps constitute indirect evi­dence for lack of interbreeding only if thepotential sexual partners have a chance tomeet. Sympatric samples do have anotheradvantage. When we note that two samplesare different , we are looking at their totalphenotype .We do not know how much ofthedifference is genetic and how much is due toenvironmental effects. If the two sampleslive at the same place, in the same environ­ment, any observed difference must, ofcourse, be genetic. This raises the very gen­eral and very serious problem of what to dowith allopatric samples (non-sympatric).Thetraditional attitude is to leave such decisionsto the wisdom of the taxo nomist (a most vari­able attribute). There is no clear rule andthis is most annoying because most Olivasamples in our collections are allopatric.

(2) Separations should rest upon unbi­ased samples. In practice, it is often diffi­cult to check if this obvious requirement ismet. The field collector, especially when deal­ing with abundant species, is often inclinedto collect only the most 'int erest ing' speci-

mens (extremes in the distribution of somefeature),neglecting the 'uninterest ing' inter­mediates. Such biased samples, when stud­ied by unaware researchers, are a frequentcause of taxonomic errors.

(3)The discriminating characters shouldnot be sex-related. In the case of Oliva , nosignificant sexual dimo rphism in she ll char­acters could ever be evidenced by us , so itfortunately seems that this aspect can be ig­nored in Oliva taxonomy.

(4) There is another, often overlooked riskof errors: the growth pattern of many spe­cies undergoes rather abrupt changes, solarge and small shells have a different shape.If this phenomenon is und etected , there isan obvious danger of splitting cons pecificyoung and old specimens into separate, ar­tificial "species". The risk of error can behigh, because the 'miss ing links' in size canbe quite rare (see Tursch 1997). In mostOliva species growth is approximately iso­metric (see Fig. 18): large and small shellshave similar shapes. But this is not alwaysthe case (see Fig. 19) and the juveniles ofmany Oliva species have indeed been de­scribed as separate species.

Page 17: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998 ) 279

To avoid that pitfall, one should exert spe­cial caution whenever a morphological gapis correlated to absolute size. Data on obvi­ously juvenile shells should be interpretedwith great caution and one should endeav­our to obtain complete, uninterruptedgrowth series. Non-isometric growth is of­ten betrayed by changes in the slope of thespire whorls. Conspecificity can then be es­tablished by size-invariant characters (in­tensive variables such as protoconch meas­urements) and by the analysis of colour pat­terns.

THE MORPHOLOGICALCONTINUUM

The analytical tools described here above arevery sensitive. We soon found out that theydo neatly separate a great number of sam­ples, so our first impression was that therewere a very large number of Oliva species,possibly several hundred.

It quickly became evident that this wasnot the case .We were often dealing only withdifferent populations of the same, geographi­cally variable species. There is nothing sur­prising in this: it is fully expected that dif­ferent populations of a same species couldbe separated by a suitable combination ofcharacters (see Mayr 1969 ; Futuyma 1986).

It has been shown (Tursch 1994) thatOliva species consist of a mosaic of distinctpopulations, each being quite homogeneous.When a large enough number of such local ,conspecific populations are compared, theyinvariably show considerable character over­lap. The species is thus represented by amorphological continuum in the attributehyperspace. This is a set in which no popu­lation (or group of populations) can be sepa­rated from all the others. Even if two (ormore) of the populations forming the con­tinuum can be easily separated from eachother, the gap is invariably bridged by an­other conspecific population (or a morpho­logically unbroken chain ofpopulations).Theboundaries ofthe morphological continuumare the limits of the phenetic variability of

B

Figure 20. The concept of morphological con­tinuum (see text this page). Samples A, B, C andD can be allopatric.

a species..Conversely, all the members of a morpho­

logical continuum belong to the samemorphospecies. Ifone cannot detect any gapbetween groups A and B (see Fig. 20), thenit follows that sp. A =sp. B. In the same way,sp. B = sp. C, and sp. C = sp. D. It is thenhard to escape the conclusion that sp. A =sp. D. This is a useful acquisition, becausenow we have a consistent, logical tool fortaking taxonomic decisions on allopatricsamples, a problem hitherto left to personalinterpretation. For an example of the use ofcontinua, see Tursch & Greifeneder (1996 ).It follows that interspecific discriminantsshould not anymore be searched betweenisolated populations but between continua(encompassing large sets of populations).Note that constructing a continuum is fast,because few specimens are needed to dem­onstrate overlap. Separation is another mat­ter.

THE SCALE OF SYMPATRYWe first ran into trouble during a study ofOliva oliva (Tursch et al. 1992). Two sym­patric samples (both from Phuket,Thailand)were easily separable. The samples werelarge and we had no reason to doubt the lo­cality. So they had to be considered differ­ent species. But these two samples were alsomembers of the same morphological con­tinuum, linked by an unbroken chain of

Page 18: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

280 Tropical Marine Mollusc Programme (TMMP)

(f) DURANGIT ~~: oReet !!!.

~ ~::J ::JlO lOc

~LAI NGc

3 3

Island

2

Figure 21. The scale of syntopy (see text thispage). Example: two strikingly different popu­lations of 0. oliva (Linnaeus, 1758) live in closeproximity at Hansa Bay, Papua New Guinea.Both belong to the same morphological con­tinuum (see Fig. 23).

allopatric populations with intergrading fea­tures. So they had to be the same species.This was a real problem: how could theseshells be two species in Phuket and only onespecies in the global distribution? Was theresomething wrong with the morphologicalcontinuum approach?

The same difficulty occurred again formany Oliva species from many places, butwe noticed that it happened only with shellsthat were not collected by ourselves. So theproblem was more probably with the locali­ties on the labels (or, more exactly, with theinterpretation given to the localities). Thishypothesis was checked by carefully compar­ing cons pecific samples with very accurateand detailed locality data, all collected in

Figure 22. The scale of syntopy (see text p. 19).Example: two slightly different (but fully sepa­rable) populations of 0. amethystina (Roding,1798) live in close proximity at Hansa Bay, PapuaNew Guinea. Both belong to the same morpho­logical continuum (see Fig. 24).

Hansa Bay, a small bay on the Northerncoast of Papua New Guinea, where Olivahave been under scrutiny for over 20 years.

It was found (Tursch 1994) that com­pletely separable populations of a same spe ­cies can occur within very small distances.The difference between neighbouring popu­lations is sometimes very striking, for in­stance in the case of the extremely variable(see Tursch et al . 1992 ) O. oliva (L., 1758) ,the type species of the genus, which is foun donly on exposed, dynamic beaches.Two verydifferent populations occur nearly side byside (see Fig. 21). One (A, consisting mostlyof"black" shells) lives on the long black sandSisimangum Beach , the other (B, all "white"shells) lives on the small white sand Boro

Page 19: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 281

RiD PAT17 I PAT 18

0.650.72

0 .60

0 .6 8

0 .55

0 .640 .50

0 .45 0 .60

0 .40UH 0 .56

1.2 1.3 1.4 0.58 0 .62 0.66

Figure 23. Variability of populations of 0. oliva(Linnaeus, 1758) (see text this page ). Scatter dia­gram of D/L vs. pat15/pat16. Minimum convexpolygon s. Two population s from Hansa Ba y,Papua New Guinea (A: "white" shells from BoroBeach ;B: mixture of"black"(common) and "white"(rare) shells from Sisimangum Beach) are sepa­rated by an obvious morphological gap. This gapis bridged by other conspecific populations (C:she lls from Samarai, Papua New Guinea; D:she lls from Phuket; Th ailand, E: shells fromZamboanga , Philippines) (modified from Tursch1994).

Beach. The two populations are readily sepa­rable and the separation is not correlated toshell colour (sample A contains both "black"and "whi te" shells). They are the samemorphospecies because they belong to thesame morphological continuum (see Fig. 23).

The difference between neighbouringpopulations can also be very small, (but stilldetectable by sensitive methods).This is thecase for O. amethystina (Widing, 1798 ), aspecies strictly confined to coral sand mixedwith coral rubble (in which the shell is verywell camouflaged). Here again, two separa­ble populations occur within a short distance(see Fig. 22). One (A) lives around a smallwreck C'Davit 's wreck").The other (B) is dis­tributed around Laing Island and DurangitReef Here again, the two populations are

Figure 24. Variability of populations of Oli vaam ethystina (Roding, 1798) (see text this page).Scatter diagram ofD/L vs. pat17/pat18. Minimumconvex polygons. Two populations from HansaBay, Papua New Guinea (A: shells from "Davit'swreck"; B: shells from Durangit Reef) are sepa­rated by an obvious morphological gap. This gapis bridged by other conspecific populations (C:shells from "Mast wreck",Hansa Bay, Papua NewGuinea; D: shells from Solomon Is.) (modifiedfrom Tursch 1994).

part of one same morphological continuum(see Fig. 24). Interestingly, one of the habi­tats is the coral sand surrounding a datedWorld War II wreck, isolated in a zone ofdark muddy sand (unsuited for O. amethy­st ina ). This afforded a clue to the tempo ofmorphological changes in Oliva populations(for details and discussion of possible causes,see Tursch 1994).

Such morphological divergence of popu­lations is not a local phenomenon: collect­ing trips soon detected similar separationsof neighbouring conspecific Oliva popu­lations in many places around the world. Soit ap pears that the Oliva species are a mo­saic of populations, fairly reflecting discon­tinuities in the habitat. Some populationsare very widespread, others are very re­stricted, according to the size of continuous

Page 20: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

282 Tropical Marine Mollusc Programme (TMMP)

tracts of biotope. It follows that, in manyspecies of the genus Oliva, the notion ofsympatry must be reduced to syntopy (in themeaning of 'st rictly living together, in thesame microhabitat'). Conversely, if twoclosely related Oliva forms are never foundtogether, this should be taken as a strongindication (not a proofl ) that they might beonly different populations of the same spe­cies (Tursch 1995).We should then look hardfor other populations with bridging charac­ters.

We can now understand what happenedwith the Phuket samples that caused somuch worry, as mentioned above. There aremany disjunct beaches in Phuket, and ifoursamples came from different beaches theywere not sympatric but allopatric. So therewas nothing wrong with the morphologicalcontinuum approach. We simply had no rightto interpret the gap separating the samplesin terms of species.

EPILOGUEIn conclusion, it appears that Oliva are notmore difficult to classify than many othermollusc groups, provided one selects taxo­nomic characters that are operational andanalyses them with an appropriate meth­odology. What we first called the 'good spe­cies' (in Section 1) are the species with littleor no geographic variation. In the 'bad spe­cies' most populations are different, evenover very small distances. Of course, if weneglect the existence of morphological con­tinua and start naming the differentpopulations, we might end up with as manyunfounded 'species' as there are habitats.This is exactly what did happen in the un­fortunate history of Oliva taxonomy.

In an apparent paradox, we are now backto using the criteria of colours and colourpatterns that we have so much disparagedpreviously. But we use them only for quickidentification clues (the separation of thespecies was demonstrated on other, safercharacters).

With the use of proper tools, the number

of Oliva species has decreased very muchand is still decreasing. New collecting tech­niques (SCUBA diving, deep sea trawling)have yielded many interesting deep-waterforms of known taxa but failed to deliver theprofusion of new species expected by some.This is not surprising because the greatmajority of Oliva species are found between1 and 10 m depths (see Van Osselaer, Bouil­lon, Ouin & Tursch 1993).There are presum­ably a few species (probably of small size)still lurking around, awaiting to be discov­ered. The chances of finding them seem bet­ter in rarely visited waters, such as muddybottoms with little appeal for tourists. Ex­ploration of such places could undoubtedlyyield as many undescribed forms as one canwish for - but should these really be de­scribed? Chances of finding novel speciesmight be higher in the drawers of old collec­tions. Our team has described only two spe­cies (Tursch & Greifeneder 1989; Tursch &Greifeneder 1996) and both had alreadybeen collected by Hugh Cuming before 1800.

The time now seems ripe for directingmore efforts towards a more promising (andstill largely unknown) topic : the natural his­tory of Oliva species.

ACKNOWLEDGEMENTSI am most grateful to all the friends whohave made (or still make) Oliva researchsuch a pleasant venture: Ralph Duchamps,Luc Germain, Dietmar Greifeneder, Domi­nique Huart, Sidney Johnson, Yuri Kantor,Yonous Machbaete, the late Millar Magap,Olivier Missa, Jean-Marc Ouin, Jean Pierretand Christian Van Osselaer. I thank Dr R.N.Kilburn for helpful comments on the manu­script. I am indebted to the Fonds Nationalde la Recherche Scientifique and to BIOTEC,S.A. for unfailing support.

REFERENCESAbbott, R.T. 1991.Those puzzling Olive shells. ­

American Conchologist 19(3): 23.Futuyma, D.J. 1986. Evolutionary Biology. 2nd

edition. Sinance Associates, Sunderland,

Page 21: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

Phuket Marine Biological Center Sp ecial Publication 18(2): 263 -284 . (1998) 283

Mass.Gould, S.J . 1984. Covariance sets and or dered

geographic variation in Cerion from Aruba,Bonaire and Curacao : a way of stu dy ingnonadaptation. - Systematic Zoology 33 (2):217-237.

Kohn,A.J. 1997. Infaunal sand-dwelling molluscsat Hila, Ambon, Indonesia. - Abstracts of the8t h Tropical Marine Mollusc ProgrammeWork shop/Congress: 49.

Mayr, E. 1969. Principles of sys te matic zoology.McGraw Hill, New York.

Panchen ,A.L. 1992. Classificat ion, evolution andthe nature of biology. - Cambridge UniversityPress.

Sneath, P.H. & R.R. Sokal. 1973 . Numerical tax­onomy.W.H. Freeman and Co., San Francisco.

Tursch, B. 1992 . Le Desordre des Oliva , suivi d'unElo ge de la Methode. - Apex (special issue):21-28.

Tursch , B. 1994. Th e scale of sympatry in thegenus Oliva. (Studies on Olividae 21). - Apex9(4): 131- 142 .

Tursch, B. 1995 . Micro-endemism, allotopy andtaxonomy in the genus Oliva. Abstracts ofthe12th International Malacological Congress,Vigo: 107-109.

Tursch, B. 1997. Non-isometric growth and prob­lems of species delimitation in the genusOliva. (Studies on Olividae 27). - Apex 12(2­3): 93-100.

Tursch , B. & J . Bouillon. 1993. Fasciolar regionmeasurements as taxonomic characters in thegenus Oliva. (Studies on Olividae 16). - Apex8(1-2): 11-18.

Tursch , B. & L. Germain. 1985 . A morphometricapproach to the Oliva problem. (Studies onOlividae 1). - Indo-Malayan Zoology 3: 331­352.

Tursch , B. & L. Germain. 1986 . Further proto­conch morphometrical data for Oliva tax­onomy. (Stu dies on Olividae 2). - Apex 1(2):39-45.

Tursch , B. & L. Germain. 1987 . Fi ve addit ionalprot oconch characters for Oli va taxonomy.(Studies on Olividae 5). - Apex 2: 59-68.

Tursch, B., L. Germain & D. Gr eifeneder. 1986.Oliva annulata Gmelin, 1791 (of authors): a

confusion of sp ecie s. (Studies on Olividae 4).­Indo-Malayan Zoology 3: 189-216.

Tursch , B. & D. Gr eifeneder. 1989. Oliva chryso­plecta sp. n., a familiar, undescribed WesternPacific species. (Studies on Olividae 11). -Apex4(4): 69-84.

Tursch, B. & D. Greifeneder. 1996. The "Olivaminiacea comp lex", with the desc ription of afamiliar, unnamed species. (Stu dies on Olivi­dae 25). Apex 11(1): 1-49.

Tursch, B., O. Missa & J. Bouillon. 1992 .Th e taxo­nomic structure of Oliva oliva (auct. ). (Stud­ies on Olividae 14). - Apex 7(1): 3-22.

Tursch, B., J .M. Ouin & J. Bouillon. 1995 . On thestructure of a population of Oliva oliva (L.,1758) in Papua New Guinea . (Studies on Olivi ­dae 22). - Apex 10(2-3): 29-38 .

Tursch, B. & C.Van Osselaer. 1987 . Suture meas­urements as taxonomic characters in the ge­nus Oliva. (Studies on Olividae. 6). - Apex 2(3­4): 69-84.

Van Osselaer, C., J . Bouillon , J .M. Ouin & B.Tursch. 1993. The distribution of Oliva spe­cies and the vari ation of their colour patternsin Hansa Bay (Papua New Guinea). (Studieson Olividae 18). - Apex 9(2-3): 29-46.

Van Osselaer, C., J . Bouillon & B. Tursch. 1993.Data on depth of burrowing, motion andsubstrate choice of some Oliva species. (Stud­ies on Olividae 17). - Apex 8(4): 151-158.

Van Oss ela er, C. & B.Tursch . 1988.Ten additionalsuture characters for Oliva taxonomy. (Stu d­ies on Oliv idae 9). - Apex 3(4): 81-87 .

Van Osselaer, C. & B.Tursch .1993.Anterior notchmeasurements as taxonomic characters in thegenus Oliva. (Studies on Olividae 15). - Apex8(1-2): 1-10 .

ISSN 0858-3633

Page 22: TAXONOMIC PROBLEMS IN THE GENUS OLIVA · 2018-02-07 · Phuket Marine Biological Center Special Publication 18(2): 263-284. (1998) 263 TAXONOMIC PROBLEMS IN THE GENUS OLIVA BernardTursch

284 Tropical Marine Mollu sc Programme (TMMP)