A review of the occurrence of Corbicula in the Pleistocene of North-west Europe

¹ Corresponding author. Netherlands Institute for Applied Geosciences TNO, P.O. Box 80015, 3508TA UTRECHT, The Netherlands; e-mail: t.meijer@nitg.tno.nl

² Department of Zoology, University of Cambridge, Downing Street, CAMBRIDGE CB2 3EJ United Kingdom; e-mail: rcp1001@cus.cam.ac.uk

The bivalve Corbicula fluminalis is perhaps the most famous interglacial mollusc known from north-west Europe. In England it was first described (as Cyrena trigonula) from Stutton, Suffolk, by S.V. Wood as early as 1834 and it was subsequently discovered as fossils in France, Belgium, The Netherlands, Germany, Austria and Denmark and as far east as Omsk (Kennard & Woodward, 1922; Von Linstow, 1922). Corbicula has a distinctly southern modern range, occurring in Greece and Sicily (C. gemmellaeii Philippi), and extending from north-east Africa through Syria and Asia Minor to Kashmir (Ellis, 1978). Its occurrence in Pleistocene interglacial deposits of NW Europe was therefore taken to indicate periods when summer temperatures were warmer than those of today. Despite being such a well known Pleistocene fossil, a number of important questions about Corbicula have never been satisfactorily answered. The first concerns its exact identity and whether the Pleistocene fossils are really conspecific with Corbicula fluminalis (Müller, 1774) described from the River Euphrates. The second, obviously related to the first, concerns its ecology and the palaeoclimatic inferences that can legitimately been drawn from its presence and whether it can tolerate mildly brackish water. The third question concerns its geological history and whether its occurrence can be used to distinguish interglacials of different ages. The aim of the present paper is to review the occurrence of Corbicula in the Pleistocene of NW Europe and to address each of these questions.

Photographs of a co-type of Müller's Tellina fluminalis from the River Euphrates, preserved in the Zoological Museum of the University of Copenhagen, have been published by Kennard & Woodward (1926, pl. 9, figs 4a-d). These show a rather narrow shell with prominent umbos which can be matched exactly in form and sculpture with examples of Corbicula from the European Pleistocene. The British shells show great variation, sometimes being rather thin and compressed, with relatively small umbos and a broad shell; at other times they may be thick and robust, tumid, with strong, prominent umbos and a shell that is more trigonal and taller in proportion to its breadth
Thicker shells tend to also to have a stronger, more regular sculpture. Such variation is probably controlled largely by environment. The flatter, broader shells, which may attain a large size, are mainly found in fine sands and silts, whereas the thicker shells tend to occur in the coarser, more gravelly deposits of open rivers. However, considerable variation is commonly found among series of shells from a single site.

Connolly (1930, p. 43) considered that modern Corbicula fluminalis from the Euphrates differed specifically from the commonest form of Corbicula living in the River Nile, to which he applied the name C. consobrina (Cailliaud, 1823). He claimed that C. fluminalis sensu stricto was characterized by its narrow shell, deflected umbos, and relatively coarse sculpture, whilst C. consobrina could be distinguished by its broader shell, less prominent umbos and finer sculpture. Connolly observed, however, that true C. fluminalis probably occurred in Egypt under the name of C. artini Pallary, 1902. Gardner (1932), in a comprehensive study of the recent and fossil Mollusca from the Fayum Depression, figured series of both C. consobrina and C. artini from that area. These figures show that intergradation takes place between the two taxa and that intermediate forms occur (pl. 7, figs 23-4). Despite this she retained the trivial names consobrina and artini for the two extremes. Gardner observed that English Pleistocene Corbicula are comparable to C. consobrina, rather than to C. artini, but she was apparently familiar only with shells from the Cambridge area. Baden-Powell (1934) subsequently applied the name C. consobrina to English Pleistocene material in general.

This usage does not appear to be well founded. In most large series of Corbicula from the English Pleistocene a complete intergradation occurs, as in Egypt, between tall, thickened artini or fluminalis types and broader, thinner consobrina types. Similar variation is typical of other freshwater bivalves, such as members of the genus Pisidium and Sphaerium, and is controlled by the nature of the environment.

Laville (1898) similarly noted that specimens of Corbicula from Cergy (France) can be divided into two groups, one with thin shells and the other with thick. Two specimens of the latter resemble var cor Lamarck, as do specimens from the March gravels mentioned by Dollfus (1884).

Many authors therefore consider C. consobrina to be only a subspecies or race of the Asiatic C. fluminalis and some populations from the Lower Nile and Lake Tana are morphologically inseparable. However, according to Van Damme (1984), most Sudanian populations do differ from C. fluminalis s.s. but show much variation, which might indicate the existence of several strains. Van Damme retained the name consobrina for the Sudanian species, noting the existence of four fundamental forms that are linked through a series of intermediates, possibly just reflecting ecophenotypic variation.

Populations have rarely been studied within the natural range of the genus. Morton (1986), after consideration of ecological, physiological, population and reproductive data from China and Japan, reaffirmed that only two species of Corbicula (fluminea and fluminalis) occurred throughout Asia. According to Britton & Morton (1986) the two taxa are difficult to distinguish by shell morphology when young but adult C. fluminalis (> 1 year) bear more deeply impressed growth-lines and more widely spaced sulcations on the shell exterior and tends to be taller (less elongate) than C. fluminea.

Recently, Corbicula was introduced to The Netherlands (Blanken, 1990). Initially only one form seemed to be present, but later two distinctive forms were distinguished, which spread more or less simultaneously. Both forms are now considered to be separate species, although initially confusion existed about the correct names that should be applied. Kinzelbach (1991) called them C. fluminea and C. fluviatilis but Moolenbeek (1993) subsequently demonstrated that the proper names for these taxa should be C. fluminea and C. fluminalis. These recent species are probably not conspecific with the Quaternary fossil C. fluminalis which is, as stated above, also polymorphic. Meijer (in Gittenberger et al, 1998) observed a large difference in the sculpture of the juvenile shell of fossils and the two recent species in the Netherlands. Juvenile shells of C. fluminea, as well as C. fluminalis, are far more coarsely ribbed than any fossil Corbicula shell known from the Netherlands. For this reason Gittenberger et al. (1998) consider both forms of the fossil species as a distinct taxon, the status of which is as yet unknown. However, the present authors consider that the taxonomic status of fossil, as well as recent species, remains unresolved. The taxonomy cannot be resolved within the scope of this limited review, so the widely used name C. fluminalis will be retained for the European Quaternary fossils.

Corbicula is a notorious colonizer that can occur in such numbers as to be a pest of potable and industrial water supply systems. The genus has spread widely through the waterways of North America and many ecological studies, aimed primarily at controlling it, have been undertaken in recent years (e.g. Britton & Morton, 1979). Although absent from the Holocene in NW Europe, the genus has recently been introduced into this region (Mouthon, 1981). In The Netherlands, for example, it invaded many Dutch rivers within only a four year period (Blanken, 1990; Kinzelbach, 1991). Den Hartog et al. (1992) considered the reasons why this invasion of the Rhine had been so successful. They pointed out that both Corbicula, and the amphipod Corophium curvispinum that invaded the Rhine at the same time, are r-strategists and are consequently able to respond opportunistically when circumstances become favourable. The properties of rapid growth, production of several generations per year, early maturity and considerable fecundity are ideal attributes for a successful colonizing species.

Presumably C. fluminalis would have behaved in a similar fashion during the Pleistocene, invading and spreading rapidly through river systems following initial colonization. What is clear is that Corbicula can live both in fluvial sands and gravels in relatively high-energy environments and in silts and muds in quieter habitats, the variation in shell morphology probably reflecting these environmental differences. In Africa the genus is known from some of the larger lakes, such as Lake Chad (Van Damme, 1984).

One outstanding issue is whether Corbicula can tolerate brackish water. Baden-Powell (1934) was specifically interested in this question in his paper on the March gravels in Cambridgeshire. He noted that smaller specimens of Corbicula from fluvial sediments near Cambridge, such as those from Barnwell, had a smaller height in proportion to their length than the smaller specimens from the "marine bed at Manea" (p. 207). Although he considered the possibility that the Corbicula in the March gravels had been derived "from the Cambridge district", he thought it more probable that it had "adapted itself to life with the marine forms, in spite of the cold conditions under which they lived" (p. 212). Such a conclusion now seems unlikely in view of the extent of re-working that has occurred in the Fenland (e.g. West et al., 1994). Corbicula occurs in many North Sea boreholes together with Cerastoderma glaucum and Macoma balthica, which both tolerate lowered salinities. The observation that the number of Corbicula specimens decreases when more euhaline species occur shows that the Pleistocene species was tolerant of brackish conditions (T. Meijer, unpublished data). The extant C. fluminea has been recorded in brackish sections of estuaries and is definitely salt tolerant (up to 14 ppt) and in The Netherlands occurs in greatest densities in the tidal area of the river (Den Hartog et al., 1992). Similar observations were reported from elsewhere (Mienis, 1991). Britton & Morton (1986), reviewing Asian data observe that C. fluminalis is usually found in the brackish lower reaches of river mouths (in salinities up to 24 ppt).

A second ecological issue concerns the palaeoclimatic inferences that can be drawn from the occurrence of Corbicula in the European Pleistocene. It might have been tempting, purely from its southern range, to exaggerate the contemporary mean annual temperatures but when the whole associated biota is considered, it is clear that during the middle part of the interglacial these were only 1-2° above present-day values. Observations on C. fluminea show that grow fast at 25°C (about 1 mm a week), that growth stops below 15°C, and that considerable mortality results within 1 week at 0°C, so that the population in cold water may decrease (Den Hartog et al., 1992).

In a review of molluscs from the 'Last interglacial', in SE England, Sparks (1964) pointed out the contrast between the delayed arrival of thermophilous species, such as Corbicula, at the beginning of the interglacial and their persistence towards its end, when climatic conditions show clear signs of deterioration. This situation was particularly clear at Stutton, Suffolk, where fluvial conditions remained relatively uniform but there was a steady decline in southern taxa, especially Corbicula, towards the end of the interglacial (Sparks & West, 1964). Such evidence serves as a warning against the uncritical assumption of thermal optimum conditions whenever Corbicula is found as a fossil, even in situations where derivation can be excluded.

The earliest European record of C. fluminalis appears to be a Late Pliocene record from Montagny-les-Beaune, France (Schlickum et al., 1978), where it occurred with Corbicula trigona (Tournouër). A record from the Red Crag at Waldringfield, in East Anglia (Bell, 1871), may be an error since the specimen cannot be traced. There are, however, several records from the Norwich Crag in East Anglia (Nyst (1838) described his duchasteli from Norwich Crag specimens). In the Netherlands, it is known from a number of localities ranging in age from the Tiglian to the Leerdam Interglacial (Meijer, 1987, 1988a & 1990; Spaink, 1968; Tesch, 1929a, 1929b & 1944). Corbicula is not known from the ‘Cromerian Complex’ of either England or The Netherlands. All the published records of it from the British Cromer Forest Bed Formation are based on misidentifications (Meijer & Preece, 1996).

The record of Corbicula in NW Europe after the Anglian/Elsterian is an important issue to clarify. Until relatively recently, interglacial deposits following this major glacial stage have been assigned to one of two temperate stages, namely the Holsteinian/Hoxnian or the Eemian/Ipswichian. It is clear from a consideration not only of the marine oxygen isotope record (Shackleton & Opdyke, 1973) but also from palynological records from long lacustrine sequences (Tzedakis et al., 1997), that the situation is more complex than this and that at least four interglacials preceding the Holocene have occurred since the Elsterian/Anglian. These have been assigned to oxygen isotopes stages 11, 9, 7 and substage 5e respectively although, except for the Last Interglacial (5e), these correlations are inferred rather than firmly established (cf. Sarntheim et al., 1986). Recent amino acid analyses from the Netherlands indicate that a correlation with oxygen isotope (sub)stages 9, 7c, 7a, and 5e is another possibility (Meijer & Cleveringa, in prep.).

The published records of Corbicula from interglacials after the Elsterian/Anglian therefore need to be critically evaluated. Such a comprehensive review cannot be undertaken within the constraints of the present paper but the history of C. fluminalis can be elucidated for two critical regions, southern England and The Netherlands.

Southern England

Corbicula is absent in the earliest interglacial sediments deposited by the River Thames after its diversion into its present valley during the Anglian Stage. These deposits, the so-called Lower Gravel and Lower Loam, which both contain temperate molluscs, occur at Barnfield Pit, Swanscombe, Kent (Kerney, 1971). Corbicula does occur, however, higher in the same sequence in the Middle Gravel. The sediments at Swanscombe are oxidized and are generally devoid of pollen. However, those at Clacton-on-Sea, Essex, which are believed to be contemporaneous with those at Swanscombe, are organic and have yielded pollen. This has enabled them to be related to different sub-stages of the interglacial (Turner & Kerney, 1971). As at Swanscombe, Corbicula is absent from the earliest freshwater deposits at Clacton and only occurs, as drifted valves, in the Estuarine Beds that formed during the late temperate sub-stage (Ho IIIb). Kerney (1971) used this evidence to suggest that the Lower Gravel and Lower Loam at Swanscombe formed during the early temperate sub-stage (Ho II), whereas accumulation of the Middle Gravel did not begin before the end of the late temperate sub-stage (Ho III). The deposits at both Swanscombe and Clacton are believed to have formed during OIS 11 (Bridgland, 1994), a suggestion that has recently received support from uranium series dating of authigenic carbonates at classic Hoxnian sites such as Marks Tey, Essex (Rowe et al., 1999). At a new locality 20 km SW of Clacton, at East Hyde, Tillingham, Essex, Corbicula has also been discovered in tidally influenced fluvial sediments from the later part (Ho IIIb) of this interglacial (Roe & Preece, 1995; Roe, 1999).

Deposits from four sites in the Thames system seem to have formed during the subsequent interglacial, represented at Belhus Park, Hackney Downs, Grays and Purfleet. There are two further sites in Essex: Barling, which is situated in the lower reaches of the present Thames valley, and Cudmore Grove, located to the north-east, in eastern Essex, which probably represents a tributary (Bridgland, 1995; Roe, 1999). Only a sparse vertebrate fauna is known from Barling but the vertebrate assemblages from the other sites are remarkably similar, but different in a number of important respects to those from earlier (i.e. those from OIS 11) and later deposits (Schreve, 1997). Corbicula is known from each of these sites and at Barling it occurs in tidally influenced sediments during the early part of the interglacial (Bridgland et al., in press). Such evidence reinforces the view that the sediments from these sites are different in age to those at Swanscombe and Clacton.

Discrimination of sediments belonging to the next two interglacial stages (i.e. the penultimate and the Last Interglacial) has been a source of much debate that has not yet been finally resolved. In recent years it has been noted that in Britain Corbicula never occurs in sediments containing Hippopotamus (e.g. Keen, 1990; Bridgland, 1994), although there are a few early reports claiming the contrary. It is important now to discuss each of these in turn. In the Upper Thames, where both taxa occur, separately, within the sequence underlying the Summertown-Radley Terrace, their mutual exclusion is particularly obvious (Bridgland, 1994). However, Sandford (1925) believed that both species could be found in the upper division of the Summertown-Radley sequence, later termed the Eynsham Gravel (Briggs et al. 1985). Bridgland noted that supposed Eynsham Gravels yielding Corbicula never produced Hippopotamus, with the exception of Radley, where the bivalve was represented by a single abraded shell (Kennard & Woodward, 1924), probably reworked. He concluded that the deposits in which Corbicula was abundant were older and correlated with the lowermost channel-fill deposits of the Summertown-Radley sequence at Stanton Harcourt (Buckingham et al., 1996). Eynsham itself, where the fauna clearly comes from the uppermost Summertown-Radley deposit, is a Hippopotamus site, so the term Eynsham Gravel can still be applied to that division.

Another valley in which it has been claimed that Hippopotamus and Corbicula occur together is the Cam, in Cambridgeshire. Marr (1917) reported their occurrence in basal loam deposits of what was later referred to the 2nd Terrace of the Cam, at Barnwell and near Chesterton (Worssam & Taylor, 1969). Elsewhere the Cam 2nd Terrace has produced faunal assemblages suggestive of mixtures from sediments of different ages, probably including those from the celebrated site at Barrington, although the terrace there has not been assigned to the Geological Survey's numbered sequence. Barrington is another site from which both Hippopotamus and Corbicula are recorded, but it appears that the record of the latter is based on just three fragments (Kennard & Woodward, 1919), and as at Radley, these may well be derived. The records from the Cam were part of the basis for Tomlinson (1925) suggesting that the ‘Hippopotamus-Corbicula fauna’, subsequently assumed to typify the Ipswichian (e.g. Shotton et al., 1977), was represented in the terraces of the Warwickshire Avon, even though there the Hippopotamus occurs in deposits of Terrace 3 and the Corbicula in deposits of Terrace 4 (Maddy et al., 1991).

In the Middle Thames, Hippopotamus occurs at Trafalgar Square in central London, but Corbicula is absent in what appears to have been an entirely suitable environment (Preece, 1999). As both taxa are today commonly associated in African rivers, such as the Nile, there seems to be no obvious ecological reason for this interglacial pattern of occurrence. Many workers in Britain now believe that deposits containing Hippopotamus are different in age to those yielding Corbicula. Since Hippopotamus is regarded as an indicator species for the Last interglacial in Britain (Sutcliffe, 1964, 1995; Stuart, 1976), the Corbicula-bearing sediments are thought to be pre-Ipswichian.

Corbicula occurs at a number sites in the lower Thames that would seem to be younger than those of the first two post-Anglian interglacials, already mentioned. Several of these, such as Ilford (Uphall), Crayford and Aveley, have previously been attributed to the Last Interglacial, largely on pollen analytical evidence (Sparks, 1964; West et al., 1964; West, 1969; Gibbard, 1994). However, fluvial sequences invariably yield fragmentary pollen records, covering only parts of an interglacial, and these can be hard to interpret biostratigraphically. Not only are there taphonomic problems relating to the pollen source area but reworking can be a major problem in fluvial environments. Moreover, most of these correlations were made before the real complexity of the stratigraphical succession was fully appreciated. The basis of such correlations has therefore been questioned on a number of grounds including terrace stratigraphy (Bridgland, 1994), aminostratigraphy (Bowen et al., 1989) and vertebrate palaeontology (Sutcliffe, 1964, 1975, 1976; Schreve, 1997). This evidence suggests that these sites belong to a pre-Ipswichian interglacial, possibly equivalent to OIS 7.

There is a consensus, however, about the correlation of sites beneath the Kempton Park terrace (=Isleworth terrace), such as Trafalgar Square (Gibbard, 1994; Preece, 1999) and Brentford (Zeuner, 1959; Kerney, 1959), with the Ipswichian. Hippopotamus remains occur at both these sites but Corbicula is unknown. When evaluating the ages of particular sites it is vital not to rely solely on stratigraphical 'indicator species' because of the dangers of circular reasoning. Reworking can also be a serious problem, especially in fluvial contexts. For example, Hippopotamus remains occur, not infrequently, in all Devensian gravel aggradations of the Thames (Gibbard, 1994) and Corbicula shells are also frequently derived (cf. West et al., 1994).

Corbicula is a common fossil in Thames deposits but its occurrence is sporadic and not ubiquitous. If Gibbard is correct and all the interglacial deposits, except Swanscombe, in and east of London are Ipswichian then this fact is awkward to explain. Since Corbicula is such an invasive species, its occurrence at nearby sites in the same valley would be expected, especially since they all represent similar fluvial facies. Its absence at sites like Trafalgar Square does appear to be significant, given the large number of shells analysed (Preece, 1999). Other species of mollusc have similar geological histories to Corbicula in the Thames system. Pisidium clessini (= astartoides auctt.), for example, also occurs in the later part of the Swanscombe aggradation (Kerney, 1971) and in interglacial deposits attributed by Bridgland to two different post-Hoxnian/pre-Ipswichian interglacials but not in those of substage 5e. This evidence supports the view that Trafalgar Square belongs to a different interglacial to the other Corbicula-bearing sediments in the lower Thames (Preece, 1995).

Assuming Bridgland's (1994) succession to be correct, the conclusion is that in the River Thames Corbicula was present in each of the first three interglacials after the Anglian (i.e. those tentatively correlated with OIS 11, 9 and 7) but it was absent in the Ipswichian (OIS 5e) and the Holocene. If this situation can be shown to be true of other regions, then the occurrence of Corbicula may have important biostratigraphical significance for differentiating Last Interglacial sites from earlier ones.

The Netherlands

De Jong & Zagwijn (1983) and De Jong (1988) suggested the presence of Eemian sediments in the Schouwen Formation on the basis of palynology. However, De Jong & Zagwijn (1983), contrary to De Jong (1988), found a chaotic ‘succession’ of pollen zones, indicating reworking of the entire deposit, probably during a late phase of the Eemian or during the Weichselian. This led to the conclusion that all marine deposits in the Schouwen Formation were reworked during the Weichselian (De Gans & De Groot, 1995a & b), a conclusion that is certainly incorrect. Extensive reworking of the Schouwen Formation has occurred, probably during the Weichselian, but undisturbed estuarine and shallow marine deposits belonging to this Formation also occur.

Several aspects of the molluscan evidence of the Schouwen Formation have always been under-valued, especially the presence of certain non-marine species in the marine assemblages. The occurrence of two fluvial prosobranch species seems to be especially important. Theodoxus fluviatilis has its first occurrence in NW Europe in the Eemian (Meijer, 1988b), and is common in the marine deposits of the Eemian type area, where it has been transported by the River Rhine. T. fluviatilis has never been found together with Corbicula nor with any fluvial species, such as Pisidium clessini, thought to characterize pre-Eemian interglacial deposits. A related species, Theodoxus danubialis (=cantianus of British authors), often mistakenly identified as T. fluviatilis, has never been found in the Eemian of NW Europe. It is known from several Middle Pleistocene sites, including Swanscombe and Tillingham near East Hyde (England), Herzeele (France), Izenberge, Beveren, Gyverinckhove and Vinkem (Belgium), Berlin, Bilzingsleben and Mosbach (Germany) (Meijer, 1988b; Meijer & Preece, 1995). At Zelzate, in Belgium close to the Dutch border, Janssen (1965) found T. fluviatilis in Schouwen deposits associated with several ‘characteristic’ marine Eemian species, such as Venerupis aurea senescens and Bittium reticulatum, together with Corbicula fluminalis . Zagwijn & Paepe (1968) demonstrated the presence of an interglacial soil, which they regarded as the Eemian Rocourt soil, in relation to the marine beds. Meijer (1969) showed that the T. fluviatilis was, in fact, T. danubialis, which gave rise to serious doubts about the stratigraphical interpretation of, at least part of, this sequence. In a neighbouring Dutch borehole (54A34-Bakkersdam), T. danubialis and Corbicula were associated, together with Eemian, Late Pliocene and Eocene marine species in a deposit probably reworked during the Weichselian. In other sites in the Flemish Valley, Belgian authors reported the regular association of T. fluviatilis and Corbicula (de Breuck, de Moor & Marechal, 1970; de Moor & Heyse, 1975) but the identification of T. fluviatilis is highly questionable and has not been verified.

At borehole 37C554-Zuurland the Schouwen Formation could be divided into several units (Van Kolfschoten & De Boer, 1988). The upper part was assigned to the Eemian on the basis of palynology (De Jong, 1988) and molluscs (Meijer, 1988a: zones D-E). Several ‘typical’ marine Eemian molluscan species are present, as well as Theodoxus fluviatilis. Corbicula also occurs, but only as abraded shells, suggesting reworking. In the lower part of the Schouwen Formation, below the marine Eemian deposits, a coarse-grained sediment with a basal gravel occurs (molluscan zone F). This is underlain by finer grained sediments yielding a species-poor marine assemblage lacking typical Eemian species (molluscan zone G) but containing well preserved specimens of Corbicula and Pisidium clessini, indicating a pre-Eemian age. It therefore appears that two interglacials are present in this borehole through the Schouwen Formation: the Eemian above and a Middle Pleistocene interglacial below. A similar situation was found in a nearby borehole (37E486-Oude Leede). Outside the depositional area of the Schouwen Formation, Corbicula is known from several other sites, including a borehole at Noorderhoeve (19E117), while in the Eemian type area, Corbicula was found in a Middle Pleistocene fluvial assemblage below Saalian till. Above the till a characteristic marine Eemian assemblage was recovered in which several non-marine species were present but not Corbicula.

In the Belvédère pit, near Maastricht, interglacial deposits yielding a Middle Palaeolithic industry are present on top of the Caberg-3 terrace of the River Maas (Van Kolfschoten & Roebroeks, 1985; Vandenberghe et al., 1993; Van den Berg, 1996). All deposits were devoid of pollen, but rich assemblages of vertebrates and molluscs are present. Corbicula occurs at various sites in the pit, more or less at the same level as the interglacial (Meijer, 1985). Thermoluminiscence dates on burnt flint provided an age of 250 ± 20 ka, which was supported by an ESR date on shells of 220 ± 40 ka. Amino acid analyses on Corbicula gave a mean D/L ratio of 0.388 ± 0.014 (TABLE 1). Vertebrates indicate a post-Holsteinian and pre-Eemian age (Van Kolfschoten, 1985 & 1993). Indeed, Van Kolfschoten et al. (1993) consider the interglacial to be intra-Saalian, most probably a correlative of the Early Saalian Hoogeveen Interstadial, which they consider to be equivalent to OIS 7.

The sand pit of Fransche Kamp near Wageningen is in the ice-pushed area of the Veluwe (Ruegg, 1991). In this pit ice-pushed Middle Pleistocene deposits yielded molluscan assemblages, including Corbicula at several levels (Meijer, 1991). Pollen analyses of clayey sediments indicated a Middle Pleistocene interglacial but this was difficult to assign, although the presence of a prominent peak of Tilia may be significant. The vertebrates suggested correlation with the Belvédère Interglacial and to an age older than a micromammal assemblage recovered from a nearby pit, Leccius de Ridder, at Rhenen. Interglacial deposits here, also pushed by Saalian ice (Ruegg & Zandstra, 1981), have yielded a micromammal and non-marine molluscan assemblage with Corbicula, belonging to a pre-Eemian temperate stage (Van Kolfschoten, 1981; Meijer, 1991). Bates (1993) published a mean D/L ratio of 0.500 ± 0.211 based on Corbicula from Fransche Kamp, which included one very high ratio (0.862). Omission of this anomalous outlier would yield a mean D/L ratio of 0.379 ± 0.03, which is very close to the Belvédère value.

Shell-bearing Holsteinian deposits are rare in the Netherlands and are known only from pits in the ice-pushed hill of Neede and from a neighbouring borehole (34B217-Gelselaar). Both sites provided pollen, mammalian fossils and fluvial molluscan assemblages, the latter including Viviparus diluvianus, Valvata naticina, Parafossarulus crassitesta, Belgrandia marginata, Pisidium clessini but lacking Corbicula.

Combined evidence from Belvédère, Fransche Kamp and Rhenen points to the existence of two temperate periods before the Saalian ice advance but post-dating the Holsteinian of Neede. These periods are of interglacial status and can tentatively be correlated with the intra-Saalian Hoogeveen and Bantega interstadials.
Amino acid data (TABLE 1) suggest a correlation of the Corbicula-bearing beds in the Schouwen Formation with those of 19E117-Noorderhoeve, that occur below Saalian till. These ratios are considerably lower than those obtained from Belvédère and Fransche Kamp. It would seem that the Corbicula beds from the Schouwen Formation and Noorderhoeve can be correlated those of Rhenen. Corbicula therefore seems to have been present in two pre-Eemian/post-Holsteinian interglacials, a conclusion essentially reached over fifty years ago (Tesch, 1943) and not fully appreciated by subsequent workers.

The presence of Corbicula in the Eemian type area has already been mentioned. In addition to its occurrence in securely pre-Eemian contexts in borehole 19E117-Noorderhoeve, it has been recovered from marine Eemian deposits from several other boreholes in the same area. In these instances, the Corbicula always have a different preservation to the associated marine shells. For example, in borehole 24E344-Zunderdorp, an abraded shell of Corbicula occurred alongside a well preserved marine molluscan assemblage, confirmed by amino acid data as Eemian (Miller & Mangerud, 1985). Its occurrence in another borehole (19C648-Castricum) in the Eemian type-area is, however, harder to explain. Eemian sediments, identified on the basis of their stratigraphical context and their molluscan content, occur in this borehole between 32-43 m below Dutch Ordnance Datum. They are overlain by Weichselian fluvial and wind-blown sediments, which are in turn buried beneath shallow marine deposits of Holocene age. Saalian till occurs beneath the Eemian sediments at 43 m, so the general sequence is typical of many boreholes from the western part of the Netherlands.

Miller & Mangerud (1985) have already published amino acid data from this borehole but additional measurements are now available from two levels within the Eemian deposits: 34.00-35.50 m and 41-42 m below Ordnance Datum. The upper level yielded a well preserved molluscan assemblages, consisting of marine species of boreo-temperate affinity and temperate species of non-marine molluscs including Theodoxus fluviatilis. Also present were a large number of warm-temperate marine species of Lusitanian-Mediterranean affinity, that were thought to characterize the Eemian. These specimens, which occurred together with abraded valves of Corbicula, however, are all extremely badly preserved. Between 41 and 42 m a well preserved assemblage of Ostrea edulis occurs, containing Lusitanian species characteristic of the Eemian but no non-marine taxa or poorly preserved marine molluscs were present. The following amino acid ratios are available from this Castricum borehole (for an explanation see table 1):

34-35.50 m.: Macoma balthica (well preserved): 0.162 (± 0.016) [4]      (AAL-4223)   Macoma balthica (badly preserved): 0.216 (± 0.060) [1]      (AAL-4224)   Corbicula fluminalis (badly preserved): 0.244 [3]      (AAL-4429)   Corbicula fluminalis (badly preserved): 0.210 (± 0.030) [3]      (AAL-4429)   41-42 m.: Macoma balthica (well preserved): 0.165 (± 0.020) [4]      (AAL-4227)

The ratios of the well preserved material are consistent with Eemian ratios for Macoma from the Netherlands, whereas the ratios of the badly preserved material point to a pre-Eemian age (Meijer & Cleveringa, in prep.). These results can easily be explained by reworking of pre-Eemian material into the sample from 34-35.50 m. However, this means that many warm-temperate marine species that were previously thought to be characteristic of Eemian deposits in the North Sea Basin also occur in earlier interglacial deposits. Judging from the amino acid ratios this interglacial is identical to that recognized in the Schouwen Formation (TABLE 1).

In addition to sites on land, Corbicula is known from a large number of offshore boreholes and boxcore samples from the Southern North Sea

The distribution pattern of sites in the Southern North Sea needs some explanation. The pattern obviously reflects, in part, the geological mapping programme of the NITG-TNO, so the blank regions beyond the Belgian, English and French coasts provide no evidence for the presence or absence of Corbicula. On the other hand, the almost straight east-west northern limit of the Corbicula area off the Dutch coast does seem to be a genuine feature. North of this boundary only a few scattered North Sea records of Corbicula are known. Corbicula is commonly washed ashore on North Sea beaches of SW Netherlands (Van Regteren-Altena, 1937; Janssen et al, 1984; Moraal, 1961) and eastern England. Much of this material can now be considered to be reworked from Middle Pleistocene deposits.

Amino acid racemization data provide independent evidence that can establish the relative ages of particular deposits. Many D/L ratios from the study area have already been published and further data from 200 samples from 61 sites in the Netherlands will be published elsewhere (Meijer & Cleveringa, in prep.). Many of these ratios have an important bearing on the history of Corbicula (TABLE 1). Since Corbicula often occurs in estuarine deposits, ratios from the shallow marine, euryhaline bivalve Macoma balthica have also been included.

Three groups of ratios can be distinguished. Ratios from M. balthica recovered from Dutch Eemian deposits vary between 0.112 and 0.200. Molluscan assemblages with similar ratios all lack Corbicula. The data from the borehole at 15H57-Creil, in the Eemian type area of the central Netherlands, are typical. Both the molluscs and the palynology unambiguously indicate an Eemian age. Corbicula is absent. The ratio from well preserved Macoma shells from a shallow marine assemblage was amongst the highest yet obtained from a Dutch Eemian deposit (TABLE 1).

The second well defined group of ratios come from marine assemblages in which both Corbicula and Macoma occur. Paired ratios on each species were analysed from several samples. The analysed shells from two boreholes (19E117-Noorderhoeve and 6A53-Janum) came from deposits immediately overlain by Saalian till. At Noorderhoeve a marine Eemian assemblage occurred above the till. The ratios therefore come from contexts that pre-date the Saalian ice advance into this region. At several sites south of the maximum extent of Saalian ice, a Middle Pleistocene deposit occurred, which was overlain by marine Eemian sediments. In the 37C554-Zuurland borehole, for example, sediments yielding an impoverished marine Eemian assemblage overlie fluvial or estuarine sediments of Middle Pleistocene age (Meijer, 1988a). Ratios from Macoma were obtained from both assemblages, those from the upper one were consistent with an Eemian age, whereas those from the lower were clearly older and attributable to a late Middle Pleistocene interglacial. A similar sequence is present in many boreholes in the bottom of the southern North Sea. From these, the S01-113 hole is given here as an example (TABLE 1).

The third group of ratios were obtained from specimens of Corbicula recovered from fluvial sediments relating to the Belvédère Interglacial. Comparable Corbicula ratios from marine sediments of this age are not yet known. This group is clearly older than the second group of late Middle Pleistocene age.

The badly preserved material from the 19C648-Castricum borehole produced ratios that were intermediate between the Eemian and the late Middle Pleistocene group, but we consider that they belong to the latter.

It therefore appears from Dutch amino acid evidence that Corbicula-bearing beds are all older than the Eemian. Moreover, the data indicate that at least two groups of ratios can be distinguished in the Dutch late Middle Pleistocene.

Conclusions and discussion

The systematics of Corbicula are urgently in need of revision but it would appear that only one polymorphic species (Corbicula fluminalis) occurred in the Pleistocene of NW Europe. It was present only during temperate stages, although its occurrence should not automatically been taken to infer thermal optimum conditions. Its absence during the early parts of several interglacial stages, when climatic conditions were optimal, can probably been attributed to migrational failure. Similarly its persistence towards the end of several interglacials shows some tolerance of slightly cooler conditions. Although principally a freshwater species, there is evidence that it could tolerate mildly brackish conditions.

Its history in NW Europe can be traced back to the Pliocene. It occurred in most of the temperate stages of the Lower Pleistocene (Tiglian - Bavelian) but there are no authentic records of it from the 'Cromerian Complex' in either Britain or the Netherlands. After the Anglian it occurred in the sediments of three successive interglacials, but it appears to have been absent in the Ipswichian in Britain and the Eemian in the Netherlands. Likewise it is unknown as a Holocene fossil in NW Europe but the genus has recently been introduced to Europe during the last 20 years. It is presently undergoing a rapid expansion of its geographical range, taking only 4 years, for example, to spread throughout much of the Rhine system. Live specimens have recently been found in Norfolk (Howlett & Baker, 1999) and its rapid spread through the inland waterways of East Anglia seems imminent, unless remedial measures are taken immediately. Corbicula possesses several important attributes for a successful colonizer and the rapidity of its recent invasion provides an insight into the probable nature of its spread during the Middle Pleistocene.

In 1964 the late Bruce Sparks published a review of the non-marine Mollusca known from the Last Interglacial in south-east England. He listed some 125 species of non-marine molluscs known from ten sites believed to date from the Last Interglacial (Ipswichian). He thought that "new work seems likely to add only in detail to our present knowledge". It now seems that most of the sites he believed to be Last interglacial do, in fact, belong to older interglacial stages. The same situation is likely to be true for many continental sites previously attributed to the Eemian. Sites with Corbicula that have been assigned to the Last interglacial deserve special scrutiny, although attribution to older stages must not rely solely on the occurrence of Corbicula, if circular reasoning is to be avoided. Only when such sites have been critically evaluated in the light of new understanding about the complexity of the interglacial record following the Anglian/Elsterian, will it be possible to gain a true picture of life during the Last Interglacial.

We thank David Bridgland, Piet Cleveringa, Philip Gibbard, John Hollin, David Keen and Danielle Schreve for discussion about points raised in this paper. David Bridgland and Gijs Peeters kindly allowed us to use figures from their work.

2. Distribution of Pleistocene Corbicula in NW Europe. Circles: Middle Pleistocene sites; Triangles: Early Pleistocene sites (Tiglian - Bavelian). Distribution outside England and The Netherlands based upon: De Breuck et al., 1970; Dollfus, 1884; Halet, F., 1933; Johansen, 1904; Laville, 1898; Von Linstov, 1922; Mania, 1973; Miller & Mangerud, 1985; De Moor, 1975; De Moor & Heyse, 1975; Mourlon, 1909; Roe, 1997.

SITE MEAN D/L RATIOS AGE   Corbicula fluminalis Macoma balthica     19C648-Castricum, 34-35.50 m.    - 0.162 ± 0.016 [4] (AAL-4223)1 E   19C648-Castricum, 41-42 m. - 0.165 ± 0.020 [4] (AAL-4227)1 E   37C554-Zuurland, 27-28 m. - 0.178 ± 0.039 [5] (ABER-1245) E   19A263-Bergen, 66-67 m. - 0.179 ± 0.017 [4] (AAL-3386)1 E   S01-113 (North Sea), 3-5 m. - 0.194 ± 0.004 [2] (ABER-1250) E   15H57-Creil, 14.66-15.35 m. - 0.198 ± 0.015 [3] (UKAL-105/107) E   19C648-Castricum, 34-35.50 m. 0.244 [1] (AAL-4815) 0.216 ± 0.060 [4] (AAL-4224)1 ? LMP   37C554-Zuurland, 30-32 m. 0.301 ± 0.012 [4] (UKAL-251) 0.245 ± 0.035 [3] (ABER-1246) LMP   6A53-Janum, 23 m. - 0.290 ± 0.024 [3] (UKAL-195) LMP   Zelzate (B), 12.30-13.80 m. 0.332 [1] (UKAL-250) 0.297 [1] (ABER-1249) LMP   Zelzate (B), 12.30-13.80 m. 0.330 ± 0.04 [4] (AAL-4428) - LMP   19E117-Noorderhoeve, 43-45 m. 0.332 ± 0.009 [2] (UKAL-248) - LMP   S01-113 (North Sea), 9-10 m. 0.303 ± 0.032 [4] (UKAL-258) 0.312 ± 0.010 [5] (ABER-1251) LMP   Belvédère 0.388 ± 0.014 [4] (ABER-1096)2 - B   Fransche Kamp 0.379 ± 0.03 [3] (ABER-1094)2 - B

TABLE 1 - Amino acid analyses. Mean D/L ratios from several critical sites in the Netherlands. Arranged according to D/L ratio of Macoma balthica. Explanation: First column: Site names with NITG site identification and depth in metres below Dutch Ordnance Datum. Second and third columns: D/L ratios with standard deviation, number of analyses and laboratory code. AAL = Colorado amino acid laboratory (INSTAAR, Boulder, USA); ABER = amino acid laboratory of the University of Wales at Aberystwyth (UK) UKAL = amino acid laboratory of the University of Wales at Cardiff (UK) 1 ) = Miller & Mangerud, 1985; 2 ) = Bates, 1993. Age: E = Eemian; LMP = late Middle Pleistocene interglacial; B = Belvédère Interglacial.