The shells of bellerophontiform molluscs are bilaterally symmetrical and coiled through one or more whorls in a single plane perpendicular to the axis of coiling, a type of coiling described as isostrophic or isometric (Fig. 1). Bellerophontiform molluscs do not form a monophyletic group, although various authors have considered all of them to be either gastropods or monoplacophorans. Runnegar & Pojeta (1985 and references therein) advocated that they represent untorted molluscs within the Class Monoplacophora. That bellerophontiform molluscs are all torted and therefore members of the Class Gastropoda was most recently proposed by Harper & Rollins (2000). Other authors regard bellerophontiform molluscs as a mixture of gastropods and monoplacophoran molluscs (Peel 1991a; Wahlman 1992; Horný 1997a,b,c,d; Frýda 1999; see also Runnegar 1996) and this point of view is followed here. Peel (1991a) referred classic monoplacophorans such as TryblidiumLindström, 1880 and the extant NeopilinaLemche, 1957; and various Lower Palaeozoic forms to a Class Tergomya, abandoning the abused term Monoplacophora. While not universally accepted (Harper & Rollins 2000), this name is employed herein. The tergomyan shell is exogastrically coiled, with the apex at the anterior and the shell expanding posteriorly. On account of torsion, a bellerophontiform mollusc interpreted as a bellerophontoidean gastropod would have had an endogastric shell, with the apex posterior and the shell coil expanding anteriorly (Fig. 1).

The mantle cavity with enclosed gills is located dorsally near the aperture in both tergomyan and gastropodan bellerophontiform molluscs. Water currents enter the mantle cavity antero-laterally and de-oxygenated water is expelled medially at a position often marked by an angulation, sinus or slit. In a tergomyan this direction of expulsion is posterior but, on account of torsion, the exhalant stream in bellerophontoidean gastropods is anterior, over the head. In bellerophontoidean gastropods with a deep slit or long series of discrete openings (tremata), the exhalant stream may be displaced antero-dorsally, or even postero-dorsally, to increase separation of inhalant and exhalant streams.

Knight (1952) proposed that bellerophontoidean gastropods possessed the same general anatomy as recent pleurotomarioidean vetigastropods, with a pair of gills located symmetrically about the medial emargination. While the number of gills present in isostrophically coiled tergomyans is less certain, these too are considered to be located dorso-laterally. Linsley (1977) observed that bellerophontoidean gastropods with narrow, laterally compressed shells tend to develop a deep and narrow slit (Fig. 1C), while in wider shells a broad medial emargination often culminates in a short slit (Fig. 1B). In some cases, the location of the inhalant currents may be indicated by antero-lateral emarginations. In all these cases the inhalant currents enter the mantle cavity antero-laterally and pass through the mantle cavity prior to exhalation. Some broad shells, however, possess a deep slit (Fig. 1A), as pointed out in species of BucaniaHall, 1847 and UndulabucaniaWahlman, 1992 by Wahlman (1992). Some of these shells may even close the deep slit anteriorly to form a trema, e.g. Bucania squamosa (Lindström, 1884) from the Silurian of Gotland (Peel 1991b).

In coiled tergomyans the water currents also enter the mantle cavity antero-laterally, passing alongside the coiled anterior apex. Shells such as CarcassonnellaHorný & Peel, 1996; SinuitinaKnight, 1945 and TelamocornuBerg-Madsen & Peel, 1994 have developed distinct umbilico-lateral sinuses for the inhalant currents (Fig. 1E). In Beyrichidiscus fecundatusLinsley & Peel, 1983 a pronounced anterior gape accomplishes the same purpose.

The supposed anatomical difference between exogastric tergomyans and the endogastric gastropods add a further parameter to the analysis of the relationship between shell morphology and predatory attacks in bellerophontiform molluscs. Differences between attack strategies in morphologically similar but oppositely oriented, tergomyans and gastropods may provide corroboration of the postulated anatomical reconstructions and their resultant systematic interpretations.

Torsion, deep retraction and the development of an operculum are all gastropod features that can be interpreted as evolutionary responses to predation pressure. Thickening of the shell, various sculpturing and narrowing of the aperture may also be considered as evolutionary antipredatory adaptations (see for instance Vermeij 1977). Other anti-predator measures in modern gastropods include camouflage, fleeing the site of the attack and the release of chemicals to dissuade attackers. Only a minor chemical trace of a predator in the water mass can provoke an antipredatory response (Weissburg et al. 2002).

Buss (1995) argued that torsion, the diagnostic character of all gastropods, came about as an antipredatory invention, as a posterior tergomyan position would render the mantle cavity unprotected when the shell was lifted during respiration. This hypothesis is especially applicable in the case of elongate tryblidiid tergomyans which lack a coiled shell, but the presence of antero-lateral inhalant sinuses and a medial postero-dorsal emargination in many coiled fossil tergomyans permits respiration without lifting of the shell above the substratum. Torsion also brings the mantle cavity to an anterior position so that all sensory receptors and the inhalant streams focus towards the direction in which the animal moves. Although gastropods and tergomyans, which clamp against the substratum for protection, do not need to withdraw defensively into the shell, the mantle cavity in gastropods provides a protective chamber for the head on withdrawal, with the foot and its mounted operculum following the head into the shell. Indeed, in classic gastropod theory (Garstang 1928) this function acting in the larval stage is considered to be the reason for torsion. The mantle cavity should be large enough to accommodate the withdrawn head–foot mass and still maintain the flow of water. The depth of the slit indicates the minimum depth of the mantle cavity, but it may be much deeper. Dzik (1981) inferred that deep attachment of the retractor muscles in bellerophontiform molluscs is also associated with a deep mantle cavity. However, the deep placement of muscle scars is mainly related to withdrawal of the soft parts into the shell, a clear defensive strategy. The disposition of the principal shell attachment muscles reflects the immediate mechanical requirements of the living molluscs and Linsley (1978) argued that clamping against the substratum was more important than deep withdrawal in a variety of shell forms with a flared tangential aperture. McNair et al. (1981) proposed that clamping was a behavioural trait of all living shelled gastropods; a planar aperture is common among snails living on hard substrata, but less common in those living on soft substrata. Peel (1993) demonstrated the different requirements of deep withdrawal and clamping in the Ordovician CarinaropsisHall, 1847.

Muscle attachment areas in coiled gastropods and tergomyans clearly need to be set sufficiently deep (abapertural) into the shell interior to withdraw the entire animal into the shell or to effectively clamp the shell down over the cephalopedal mass. This depth, however, will depend both on the size of the soft parts and also on the shape of the shell, especially the rate of expansion. The deeper the animal is withdrawn into the shell, the greater is the effort required by a predator to break back the apertural regions to secure the prey. Deep withdrawal also enhances the defensive function of the gastropod operculum (but see below), as an operculum fitting just against the apertural margin is more easily bypassed than one drawn deep into the shell interior. Deep withdrawal is potentially equally beneficial to coiled tergomyans and bellerophontoidean gastropods alike, although most of the former seem to clamp against the substratum (Horný 1965; Horný & Peel 1996). Many bellerophontoidean gastropods also clamp, but many others were capable of deep withdrawal to combat predation or combined clamping with deep withdrawal (Peel 1972, 1976, 1982, 1991b, 1993; Horný 1997b).

Stanley (1982) stressed the development of the operculum, a general feature of gastropods and a clear defensive adaptation, in connection with torsion, as only a torted and endogastrically coiled gastropod with an operculum mounted posteriorly on the top of the foot, underneath the shell in life, could readily close the shell aperture. Morphological constraints exclude the presence of an operculum in exogastrically oriented tergomyan shells (Stanley 1982). Thus, if all bellerophontiform molluscs were tergomyans, no operculum would be expected, and defensive strategies would be limited to clamping and deep withdrawal. No operculum is known from a bellerophontiform mollusc, although Dzik (1981) reconstructed the Middle Ordovician (Darriwilian) tergomyan Sinuitopsis neglecta Barrande inPerner, 1903 carrying an operculum, based on similarities with muscle arrangements in operculate hyolithids and cephalopods. Northrop (1939) claimed the presence of paired opercula in Salpingostoma inornatumNorthrop, 1939, but Peel (1972) disclaimed the record. However, opercula are usually rare fossils in the Palaeozoic (Yochelson 1979) and they are not known in many groups of undoubted gastropods.

The bellerophontiform molluscs discussed herein were located in a number of museum collections and published papers. The material includes both presumed tergomyans and gastropods, indicating that the term ‘bellerophontiform molluscs’ has only a morphological connotation. All the described shell injuries and subsequent repairs mentioned below are attributed to failed predation and not to mechanical breakage (for supportive discussion see Ebbestad & Peel 1997; Alexander & Dietl 2003, 2005).

A total of 66 specimens, of which 50 are gastropods and 16 are tergomyans, preserve shell repair. It is not possible to provide a meaningful estimate of how many specimens were examined in the search for these 66 specimens, although the number is far greater. Many were noted as novelties in the course of other studies, often in collections where the state of preservation prevented accurate quantification or even recognition of repairs.

Ten specimens (all gastropods) are from the Carboniferous, two specimens (one gastropod) are from the Devonian, seven specimens from the Silurian (all gastropods) and 47 specimens from the Ordovician (32 gastropods; 15 tergomyans). The specimens were certainly derived from a wide range of marine environments, although little information on these is generally available from museum records. Information on geological age is also similarly imprecise. Taking into account the small overall sample size, we have chosen to apply a simple morphological approach to the material as a whole.

Two or more injuries occurred in 16 specimens. Tables 1–4 summarize the data and present information on depositories; tergomyans and gastropods are presented separately. An index after the taxon name in the table refers to specimens where shell repairs have been described and figured, or are visible in a figure without being described or noted by the original author. The corresponding reference is cited in the caption. Some injuries are not visible in the primary reference, such as, for instance, for Joleaudella sphenonotus in Figure 6C which was figured by Koken & Perner (1925) and Yochelson (1963). Such cases are not indexed in the tables.

Where several injuries are present, these are numbered with roman numerals either as presented in the literature or, if previously undescribed, from the oldest to youngest injury (adaperturally) and from the right side to left side of the shell. In lateral view, left side of an isostrophic shell is that visible when the anterior is located at the left (Fig. 1). In the exogastric tergomyans the shell apex is located anteriorly, but in the endogastric gastropods the apex is posterior.

For each group, four levels of sorting are applied in the tables in order to aid the general overview. Shell morphologies and terminologies used are illustrated in Figures 1, 2. Specimens are first sorted according to shell form – broad vs. narrow. The category of Broad shells includes conchs of about equal length and width (Fig. 1A, B, D, F), while the category Narrow shells includes conchs that are about one and a half to twice as long as wide (Fig. 1C, E).

The second level of sorting is the depth of the anterior medial emargination for each shell form. Most bellerophontiform molluscs have a medial anterior sinus but its depth varies (abaperturally) between taxa; in some cases the sinus passes into a slit of variable depth abaperturally. For the purpose of this paper, medial marginal emarginations are classified as either deep or shallow. Deep emarginations penetrate more than a third of a whorl back from the aperture (Fig. 1A, C, D, F), here including the trema formed in species of Salpingostoma (Fig. 1D), while a shallow emargination reaches less than a third of a whorl back abaperturally (Fig. 1B, E).

The third level of sorting is the type of injury present. The geometry of breaks may be categorized following Alexander & Dietl (2001, 2003), here including embayments (Fig. 2A, B, D), scallops (Fig. 2C, D), tracked clefts (Fig. 2F) and divots (Fig. 2E). The injury pattern observed is best conformable with a predator peeling the shell.

The forth level of sorting is the position of the specific type of injury: Medial both when the injury straddles the medial emargination (Fig. 2A, B, D), Medial left or Medial right when the injuries encroach on either the left or right side of the medial emargination (Fig. 2I), and Lateral left and Lateral right when the injury is some distance away and clearly set apart from the medial emargination (Fig. 2C, J). Left and right sides of the isostrophic shell are in relation to the orientation of the shell in life position as described above. Size classes are indicated for each specimen, representing the size of the shell at the site of the injury.

To test the role of the depth of the medial emargination in determining the distribution of shell repairs in various shell morphologies, frequency distributions were calculated with respect to the position of injuries on the shell (medial or lateral) relative to shell form (broad or narrow). These were calculated for tergomyans (Fig. 3A), gastropods with a medial slit (Fig. 3B), gastropods with a medial sinus (Fig. 3C) and all groups combined (Fig. 3D). Table 5 summarizes the statistics for the frequency distribution presented in Figure 3. The results were tested using a chi-squared test for two samples in the statistical package PAST (Hammer et al. 2001).

Secondly, the frequency of types of injuries in tergomyans and gastropods relative to their position on the shell (medial or lateral) was calculated (Fig. 4). This frequency is isolated from shell form and type of emargination to see if it is independent of the frequencies where shell form and emargination are considered (Fig. 3). Table 6 summarizes the frequency distribution presented in Figure 4.

Finally, the frequency distribution and position (medial or lateral) of embayments and scallops across size classes were calculated (Fig. 5), to estimate the role of shell size with regard to type of repaired injuries and the position of these.

The null hypothesis assumes a difference in the calculated distributions and frequencies of injuries in the shell forms. Owing to the scarcity of divots and tracked clefts (three and two of each, respectively, all positioned laterally; Tables 1, 2), these types of injuries were omitted from the above analysis and from Tables 3–6.

The 66 specimens recorded in this study document 101 repaired injuries attributed to failed predation (Tables 1–4, Figs 2, 6). In most cases, only one repair was present in each specimen, but 16 specimens showed more than one injury. Horný (1997d) recorded as many as ten injuries in a specimen of the tergomyan S. neglecta from the Ordovician of Bohemia, Czech Republic (Fig. 2D). With five exceptions all injuries can be classified as either embayments (47 injuries, 46.5%) or scallops (49 injuries, 48.5%), with divots (three injuries, 3.0%) and tracked clefts (two injuries, 2.0%) being rare.

Bellerophontiform gastropods and tergomyans with a deep medial emargination in the form of a slit have significantly more medial injuries than lateral ones, while near-equal or equal frequencies of lateral injures and medial injuries occur in specimens with a shallow emargination (slit or sinus), suggesting random distribution of injuries around the aperture. Shell form (narrow or broad) does not exert an overall control on the distribution of injuries, except perhaps in broad explanate shells with only slight expression of the medial emargination.

It is clear that many predators seek out their gastropod prey by converging on attractive chemical signals (Weissburg et al. 2002). Conceivably the polluted exhalant water current that emerged from the medial emargination in bellerophontiform molluscs could focus the attack in the area of the slit and sinus first, causing the distribution of fractures to be concentrated here. The slightly skewed distribution toward lateral injuries in sinuate species suggests no particular targeting of the exhalant current. A better explanation is therefore that the presence of a medial emargination allows the best mechanical purchase. If these molluscs actively applied clamping as a protection against predation, the emargination would be the area most exposed to an attacker, especially with the presence of a medial slit. Probably, a combination of these factors aided the predator when directing its attack, which means that the distribution of injuries was a result of both passive selection due to structural geometry and preferentially targeting by predators towards the point of weakness (i.e. site-specific mode of attack).

Dr Radvan J. Horný, Národní Muzeum in Prague, Czech Republic, generously provided photographs and kindly shared his knowledge of Bohemian specimens. Financial support from the Swedish Science Research Council (Vetenskapsrådet) through grants to J.S. Peel and J.O.R. Ebbestad is gratefully acknowledged. A stipend to J.O.R. Ebbestad from STINT, the Swedish Foundation for Cooperation in Research and Higher Education, is also acknowledged. Constructive comments by Dr Gregory Dietl, Department of Earth and Atmospheric Sciences, Cornell University, Ithaca, USA, and an anonymous reviewer are gratefully acknowledged. The study also benefitted from comments by Dr Michal Kowaleski, Virginia Polytechnic Institute and State University, Virginia, USA, and an anonymous reviewer, at an early stage. Dr Stefan Ekman, Museum of Evolution, Uppsala University, Sweden, gave helpful suggestions on the statistics. Studies of collections by Jan Ove R. Ebbestad in the Museum für Naturkunde in Berlin and The Geological Museum, University of Copenhagen were made possible through support of the SYNTHESYS program. Kind help from the curators of the many visited collections (see captions in Table 1) is much appreciated.