Introduction

Shell-crushing predation by durophagous animals such as decapod crustaceans became an important cause of death and a pervasive component of natural selection during the course of gastropod evolution, and its fossil record has been cited as a major piece of evidence for the “Mesozoic marine revolution” and “escalation” hypotheses (Vermeij, 1977, 1983, 1987). Such predation leaves characteristic breaks on gastropod shells (represented by jagged fractures on the outer lip: = peeled breaks). The origin of the breaks in such shells (peeled shells), however, is difficult to determine, that is, whether they resulted from lethal attacks on snails (primary predation, or PP), from attacks to secondary occupants such as hermit crabs (secondary predation, or SP), or from accidental post-mortem physical destruction (e.g., Shoup, 1968; Bertness, 1981; Vermeij, 1982; LaBarbera and Merz, 1992; Kowalewski, 2002). In addition, “mistaken predation” (accidental attacks on empty shells: Walker and Yamada, 1993) leaves similar breaks, which we regard as being in the SP category. To avoid this problem, previous studies (“the percent with scars method” and “the scars per shell method”; see Alexander and Dietl, 2003) have been analyzing unsuccessful, repaired breaks (represented by jagged lines in whorls) in fossil gastropod assemblages through geologic time, based on the assumption that the incidence of unsuccessful predation correlates with the proportion of an animal population killed by predators (see Vermeij, 1982). These studies have shown that the frequency of repaired breaks, although it can be biased by several factors, was very low in the Paleozoic and early Mesozoic, and has increased since that time with an apparent peak during the latest Cretaceous (Vermeij et al., 1981; Alexander and Dietl, 2003, and many references therein). The frequency of repaired breaks in gastropods, however, relies on the frequency of attacks by predators and the vulnerability of prey: a low frequency of repaired breaks suggests either that attacks were infrequent or highly lethal, whereas a high frequency of repaired breaks suggests either that attacks were frequent or not highly lethal (Alexander and Dietl, 2003; see Leighton, 2002 for other factors to the frequency). Therefore, elucidating the effect on gastropod evolution of such predation ideally depends on dealing directly with lethally attacked shells (i.e., PP).

Vermeij (1982) was the first to quantitatively evaluate PP and SP in gastropod assemblages by calculating the number of peeled and drilled shells divided by the number of drilled shells. This procedure is based on the assumption that peeled shells with a naticid or muricid drill hole are attributed to SP, because these predatory gastropods only attack living snails. This study provides the minimum estimate of SP because hermit crabs prefer undrilled shells (Pechenik and Lewis, 2000). We here propose two new methods to quantitatively evaluate the PP and SP values more in detail, based on material from the Pliocene Shinzato Formation, Okinawa, Japan, as a case study. We show that predation on hermit crabs may have been an important agent in the fragmentation of shells in fossil gastropod assemblages much more than previously thought and the new methods provide an insight into the study of this unique biological interaction through geologic times.

Materials

The fossil gastropods discussed here were collected from an exposure of the latest Pliocene Shinzato Formation on Miyagishima islet off Okinawa Island, southern Japan (locality 87-21 in Noda, 1988; Figure 1). The strata at this site consist of bioturbated sandy mud; the exposure is about 20 m thick and contains diverse upper-bathyal mollusks (200–400 fathoms according to MacNeil, 1960). The gastropod assemblage at this site can be considered to be allochthonous in origin, irrespective of no positive sign of shell transportation by currents, because the shells may have been mixed up together by the expected activities of hermit crabs. The gastropod shells from this site we used for the analysis are stout without noticeable signs of shell fragmentation by currents and sediment compaction, and are scattered abundantly in denudated residue under the subtropical climate prevailing in this area. Additionally, the sediment matrix is easily removed from the shells, so that even the delicate traces of epi- and endobionts are clearly seen under a binocular microscope. Specimens were collected by sieving the fossil-bearing residue with a 2-mm mesh and recovering and counting all complete and fragmented shells. We omitted severely fragmented and drilled shells in our analyses, and counted all the repaired and peeled breaks that appeared in the whorls for the model method. We follow the identification of gastropods of MacNeil (1960) and Noda (1980, 1988). The specimens are all deposited in the Department of Geology, the National Science Museum, Tokyo (NSM).

Figure 1

Map showing the sample locality.

Endobiont method

Figure 2 shows a taphonomic scenario for gastro-pods. When snails die, shells remain intact (Figure 2A) or are fragmented biologically by predators such as shell-drilling gastropods (Figure 2B) or crabs (Figure 2C), then incorporated into death assemblages. After being physically fragmented or not (Figure 2D, E), dead shells become a buried assemblage directly, or they are utilized by secondary occupants such as hermit crabs with or without commensal-biont infestation (Figure 2F, G). Hermitted shells are then incorporated into the buried assemblage, sometimes with biological (Figure 2H) and physical fragmentation (Figure 2J), the former of which occurs due to crushing predation against secondary occupants. Reworked shells from the buried assemblage can be reincorporated into the taphonomic process (Figure 2K). Finally, the buried assemblage turns into a fossil assemblage through diagenesis, with or without physical fragmentation (Figure 2L, M).

Figure 2

Taphonomic history of shell fragmentation.

In this scenario, biological shell fragmentation occurs in two ways: predation against snails (PP) and against hermit crabs (SP). We are unable to distinguish PP and SP for peeled shells, because they are the same in appearance. However, trace-fossil commensals of hermit crabs are useful “markers” for detecting SP in the fossil gastropod assemblages. Here we deal with two kinds of domicile drill holes produced by spionid annelids, which are obligatorily commensals on hermit crabs. One is the ichnospecies Helicotaphrichnus commensalis that is a U-shaped burrow on the columella of gastropods, where the annelids benefit from filter feeding and receiving respiratory currents generated by the hermit crabs (Kern et al., 1974; Kern, 1979; Figure 3). The other consists of small, round drill holes found only on apical whorls (we call this an “apical drill hole”) (Williams, 2000; unpublished observation; Figure 3A, F). We follow Walker (1992) in using the terms “hermitted” and “pagurized” to denote shells occupied by hermit crabs and those with associated trace fossils or body fossils, respectively.

Figure 3

Trace fossils and shell fragmentation patterns for selected gastropods. Arrowheads indicate Helicotaphrichnus commensalis, and arrowhead with asterisk indicates “apical drill hole.” A, B. Makiyamaia sp. with H. commensalis and an “apical drill hole.” NSM PM14845. C, D. Severely broken turrid specimen with H. commensalis near original aperture. C, NSM PM14846; D, NSM PM14847. E, F. Gemmula granosa with H. commensalis on columella of newly generated aperture. E, NSM PM14848; F, NSM PM14849. G, H. Gemmula granosa with H. commensalis close to original and newly generated apertures. NSM PM14850.

Such trace fossils tell us the various taphonomic histories of gastropod shells (Walker, 1988, 1989, 1992; Seilacher, 1969). Figure 4A schematically shows the distribution of PP and SP in pagurized, unpagurized, hermitted and unhermitted shells in dead and fossil gastropod assemblages. It shows that (1) peeled breaks in pagurized shells resulted from both PP and SP; (2) peeled breaks are due only to PP in unhermitted shells. The difference in frequency of peeled shells between pagurized and unpagurized shells can be regarded as the actual amount of SP when all the unpagurized shells were unhermitted (Figure 4B). However, unpagurized shells may include a certain number of hermitted shells, and hermitted, unpagurized shells include PP and SP to the same degree as do pagurized shells, because the domicile drill holes seem not to have affected predatory preference (Figure 4C). Consequently, the difference in frequency of peeled shells between pagurized and unpagurized shells is evidently smaller than the actual SP and therefore can be regarded as the minimum estimate of SP. Figure 4C shows an example of the calculation in Gemmula cassinella.

Figure 4

A. Schematic distribution of primary (PP) and secondary predation (SP) in pagurized and unpagurized shells. B. Schematic distribution of primary (PP) and secondary predation (SP) in pagurized and unpagurized shells when unpagurized shells are all unhermitted. In this case, difference in frequency of peeled shells between pagurized and unpagurized can be regarded as actual amount of SP and can be represented by shaded area: SP = (“peeled in pagurized” – “peeled in unpagurized”) × “pagurized.” C. Example of minimum estimate of SP in Gemmula cassinella (number in parentheses, see Table 1). Unpagurized shells can be expected to include hermitted shells, so that difference in frequency of peeled shells between pagurized and unpagurized can be regarded as minimum estimate of SP (area surrounded by thick line). Areas indicated by asterisks are same in size. For G. cassinella, minimum estimate of SP can be calculated as (“0.46” – “0.20”) × “0.38.”

Table 1

Results of analyses and comparisons with previous methods for 15 gastropod species.

The frequency of pagurized specimens ranges from 0.05 (Cancellaria yonabaruensis and Euspira cf. pallida) to 0.88 (Pseudolatirus yonabaruensis) for the 15 species examined, and the pagurized shells bear peeled breaks more than do the unpagurized shells in 13 species, evidently suggesting that SP is involved (Table 1). An exception is Naticarius cf. niasensis, where the unpagurized shells bear peeled breaks more than do pagurized shells, most probably due to the small number of pagurized shells. The other exception is Hindsia solida, where the thick outer lip in adults seems to have prevented effective peeling by predators. The minimum estimate for SP by this method ranges from 0.01 (four species) to 0.35 (Gemmula granosa) (Table 1).

Helicotaphrichnus commensalis is positioned in the inner lip or columella within the shell aperture (Figure 3A), and the presence of a deep peeled-break well behind the location of H. commensalis strongly suggests that it was due to predation against a hermit crab rather than a snail, because the aperture was likely intact when the shell was hermitted (Figure 3B). H. commensalis is sometimes present on the columella of a secondary aperture formed by destruction (Figure 3E–G), suggesting that the shell was hermitted after predation against its former resident(s) – either a snail or hermit crab(s). When H. commensalis is present on the columella close to both the original and secondary aperture, it suggests that the shells were hermitted at least twice, and that shell peeling was evidently performed against hermit crabs (Figure 3G, H). Even for severely broken shells, the presence of H. commensalis on the columella near the original aperture implies that the shells were intact when they were hermitted (Figure 3C, D).

The peeled breaks among the shells that bear H. commensalis on the columellas of the original aperture only resulted from SP and their frequency can be regarded as the actual rate of SP when all the shells were hermitted. However, the fossil assemblage may contain shells buried immediately without being occupied by hermit crabs. Therefore, the actual SP frequency appears to be lower than this rate and can be regarded as a maximum value. The maximum estimate of SP by this method ranges from zero (three naticid species) to 0.60 (Trochocerithium shikoensis) (Table 1).

Model method

Our model, based on the frequency distributions of the number of repaired breaks per shell (Table 2), employs three parameters: (1) frequency that a snail encounters predation in its life span (predation intensity r); (2) lethality of predation (p); and (3) probability that a dead shell suffered SP at least once (α).

Table 2

Data for model method. Expected fractions of unpeeled specimen with x repair(s) and peeled specimen with x – 1 repair(s) are given by function SUB(x) and LET(x), respectively.

First, we treat time as discrete. This model assumes that, in the absence of predation, each snail survives for its intrinsic life span L. Our snails encounter predation with a probability a for each time interval. Here, ap and a(1 – p) give the probability that they encounter lethal and sublethal predation, respectively. The predation intensity r is defined as:

Discussion and conclusions

We employed the two methods to 12 species (H. solida, Natica sp. and N. cf. niasensis were not applicable for the endobiont method), and the incidence of SP by the model method falls within the range of SP by the endobiont method for 10 species (Table 1: E. cf. pallida and Natica aff. stellata are slightly out of this range, most probably due to the small sample size for the endobiont method), which suggests that our two methods are appropriate. However, the results of our analyses are still biased because the severely fragmented shells were not included. We assume that the SP is involved more than the results because the frequency of pagurized shell is higher in the severely fragmented shells (53%; we only examined shells with apical whorls) than in the counted shells (21%). Repeated occupation by hermit crabs and their inability in shell repair result in further accumulation of peeled breaks (SP) in dead gastropod shells. Additionally, hermitted shells (mostly surface dwellers) may have more potential for encounters with the predators than live, burrowing snails.

Our study, uncomplicated as it is, has expressed for the first time the importance of SP for the fossil gastropod assemblages and evaluated SP and PP quantitatively. In elongate shells (e.g., T. shikoensis and four turrid species), the shell peelings are largely attributed to predation on hermit crabs (SP) irrespective of the very high frequencies of shell repairs, whereas they are the reverse for four spherical naticid shells. The predation lethality ( p values in the model method) is significantly correlated with shell elongation (Figure 5 and Table 1, r² = 0:635, p < 0.005), further supporting the notion that shell elongation is an antipredatory adaptation in gastropods (e.g., Vermeij, 1987). This is evidently due to the fact that the elongate shells are less vulnerable to crab predation than the spherical shells, because of the deep retraction of the soft animal into the shell and the narrowness of the apertural opening that prevents the insertion of claws (e.g., Vermeij, 1987). Due to this effect, sublethal repaired breaks are greater in number on elongate shells than on spherical shells (Schindel et al., 1982). This is shown well in Daphnella ryukyuensis (elongate) and Natica sp. (spherical), where both were equally attacked 0.30 time during their lifetime, but the incidence of repaired breaks for the former is nearly six or seven times greater than for the latter in the “percent with scars method” and “scars per shell method,” respectively (Table 1).

Figure 5

Predation efficiency (p) and predation frequency (r for 15 gastropod species from Shinzato Formation. Data from Table 1. Abbreviations: 1, Naticalius cf. niasensis; 2, Natica sp.; 3, Natica aff. stellata; 4, Euspira cf. pallida; 5, Gemmula cassinella; 6, Daphnella ryukyuensis; 7, Coronasyrinx takabanarensis; 8, Gemmula granosa; 9, Hindsia takabanarensis; 10, Cancellaria yonabaruensis; 11, Splendrillia sp.; 12, Hindsia solida; 13, Trochocerithium shikoensis; 14, Makiyamaia coreanica; 15, Pseudolatirus yonabaruensis.

The results of this study strongly imply that the frequency of shell repair is also strongly controlled by the composition of shell morphotypes and their abundance in gastropod assemblages, and suggest reconsideration of the history of crushing predation based on shell repair. The more the number of species with elongate shells in gastropod assemblages, the higher the incidence of shell repair. The evolutionary history of gastropods shows that during the Paleozoic gastropod assemblages consisted mainly of species whose shell architectures are thought to be vulnerable to crushing predation, including uncoiling, planispiral, rounded and trochiform (i.e., “archaeogastropods”), and that the elongate shells resistant to crushing predation (i.e., many caenogastropods) have increased gradually in proportion through geologic times (Sohl, 1977; Schindel et al., 1982; Vermeij, 1987). It is therefore possible that the incidence of sublethal shell repair has increased gradually through geologic time due to the increase of elongate shells. To avoid the effects of shell architecture and to evaluate the history of crushing predation correctly, we should deal with changes in repair incidence by analyzing taxonomic groups with similar shell architectures. Such studies have been undertaken only on terebrids, turritellids, naticids, and conids (Vermeij et al., 1980; Allmon et al., 1990; Dietl and Alexander, 1998; Alexander and Dietl, 2003). These studies did not show any obvious trends in the repair frequency, which synchronize with one another.

We have not yet employed our methods rigorously to shallow marine, soft-bottom gastropod assemblages. We presume however that the SP is also dominant in such environments because hermitted shells are so common there (e.g., Simoyama, 1985). The endobiont method based on H. commensalis and the “apical drill holes” appears to be difficult to apply to shallow marine, soft-bottom gastropod assemblages because the two endobionts are usually uncommon in shallow marine soft bottoms than in deepwater bottoms (Ishikawa and Kase, unpublished data). However, the endobiont method appears to be applicable by detecting other endo- and/or epibionts common in shallow waters being useful for recognizing hermitted shells (Walker, 1992). In contrast, our model method is highly useful to evaluate the shell fragmentation even for shallow marine, soft-bottom assemblages if bulk material from outcrops unbiased by collectors is available. In this case, we can only evaluate the PP because the physical shell fragmentation is not ignorable under such high-energy environments (but see Oji et al., 2003). A preliminary analysis of shell fragmentation in two cerithiid gastropods from an intertidal sand flat of Palau shows that Clypeomorus petrosus (N = 100) and Rhinoclavis aspera (N = 53) have unexpectedly low PP values (3% and 0%, respectively) in spite of the high frequencies of peeled shells (54% and 57%, respectively). This suggests that postmortem breakage is also an important agent for shell fragmentation in shallow-water gastropod assemblages as well as in deep-sea assemblages (Ishikawa and Kase, unpublished).

The fossil record of hermit crabs (Paguroidea) dates back to the early Jurassic (Briggs et al., 1993) and the direct evidence of hermit crabs occupying gastropod shells dates from the late Cretaceous (Mertin, 1941). Crab predation on hermit crabs therefore may have occurred since the late Mesozoic. In conclusion, by applying our new methods, the predation intensity (r) and lethality of predation (p) can now be estimated as separate variables in fossil gastropod assemblages, which provides an insight into a more realistic view of the history of this unique biological interaction through deep time.