The taxonomy of extant animal species is often based on biological information that is not normally preserved in fossils. Reducing discrepancies between taxonomy based only on hard tissues and that based on other biological information to a minimal level is crucial in studies focusing on species diversity that integrate extant and fossil material. In the present study, I address this issue using morphological analysis of the endemic Ogasawara Island land snails in the genus Mandarina. I first examine pairwise differences in shell morphology among 39 populations of 15 extant species that were discriminated by differences in their reproductive organs and their phylogenetic relationships. A classification model to assess whether the observed differences in fossil shell characters were inter- or intra-specific was developed by training an artificial neural network (ANN) with the patterns of differences in shell characters among the extant species. The average probability that the trained ANN misclassifies extant forms was 1.4%. The trained ANN was applied to discriminate morphological differences among the Pleistocene-Holocene fossil samples of Mandarina luhuana that occurred in Chichijima and Minamijima. As a result, three species were identified in the samples previously referred to M. luhuana. Mandarina pallasiana, previously treated as a synonym of M. luhuana, is separated, and one new species and one new subspecies are described.
Introduction
Evolutionary biologists have debated species concepts since the 19th century. Recently, much focus has been placed on the debates regarding the species concept that emphasize reproductive isolation (e.g., the biological species concept) (Mayr, 1942, 1963, 1969, 1996), genotypic clustering (Mallet, 1995), occupation of distinct ecological niches (Van Valen, 1976), and phylogenetic relationships among lineages including the phylogenetic (Cracraft, 1989; McKitrick and Zink, 1988; Baum and Donoghue, 1995; Shaw, 1998; Wheeler and Platnick, 2000; Mishler and Theriot, 2000), cladistic (Meier and Willman, 2000), internodal (Kornet, 1993), and DNA barcoding (Tautz et al., 2002, 2003; Blaxter, 2003 Blaxter, 2004; Hebert et al., 2003) concepts. Some even argue that species are no more real than any other hierarchical grouping in the tree of life (Mishler, 1999; Hendry et al., 2000). One of the reasons for this never-ending controversy is that there are a number of evolutionary processes that distinguish populations or phenotypes from each other. However, a consensus is emerging among evolutionary biologists that data gathered from various methodologies, based on various different common properties (unbranched pattern of descent from a common ancestral population, morphological similarity, genetic similarity, shared adaptive zones or ecological niches, or shared mate-recognition systems) can be useful in delimiting distinct lineage segments as species (De Queiroz, 1998, 2005; Marshall et al., 2006).
Nevertheless, problems arise when fossil organisms are considered, because paleontologists describe species based only on morphological discontinuity. Such taxa are valid within the context of describing biological diversity within the fossil record, but great caution must be adopted in using fossil species in conjunction with the diversity of extant organisms for developing evolutionary theories because of the difficulties of knowing exactly what is represented by fossil species. Although fossil species are rough approximations of those of extant organisms, an attempt must be made to align species of fossils as closely as possible with species in corresponding groups of recent organisms.
One of the practical criteria for deciding the taxonomic status of fossil organisms is the degree of phenotypic difference observed among extant species. Differences in particular morphological characters between extant species include those associated with the intrinsic isolating mechanisms, and adaptive features associated with the ecology of the species. In addition, states of some characteristics reflect shared historical or phylogenetic constraints. If these characteristics or characteristics linked to them are also present in hard tissues, differences in such characteristics can be used as criteria to decide whether an observed fossil form should be described as a nominal species or not. This approach is applied here to the taxonomy of the fossil land snails in the genus Mandarina (Bradybaenidae, Pulmonata) in the Ogasawara Islands to test its validity and usefulness for identifying species in the fossil record.
Mandarina has undergone adaptive radiation within the islands of Ogasawara (Chiba 1999a, 2004; Davison and Chiba, 2006), and includes 15 extant and 5 extinct species. Ecology and shell morphology are clearly different among sympatric species as a result of character displacement (Chiba 1999b) or reproductive isolation by habitat segregation (Chiba, 2004). Allopatric species are generally discriminated by consistent patterns of differences in genital morphology and molecular phylogenetic relationship (Chiba, 1999a). Mating between species of Mandarina with different genital morphology plausibly is not successful or reduces fertilization success, because differences in genital morphology reflect differences in the timing of sperm transfer and the strategy of mate manipulation in bradybaenid land snails (Koene and Chiba, 2006). Shell morphology is not used as a major criterion for discriminating extant species of Mandarina, because the shell morphology of many of the allopatric species is very similar. However, if variation in some shell characteristics is associated with species differences, discrimination of species based on only shell morphology is possible by developing a classification model for these characteristics.
In the present study, I developed a classification model using an artificial neural network (ANN) to discriminate observed differences in shell characteristics in Mandarina between and within species by examining the association of variations in shell characteristics with distinction of species. Then, this model was adopted for classification of the fossil samples of Mandarina from Pleistocene and Holocene deposits in Chichijima and Minamijima.
Materials and methods
Samples
For the analysis of extant populations, 15 species of Mandarina collected from 26 sample locations were used (Table 1, Figure 1). Populations of possible hybrid origin were not used, because these populations often possess shell morphology unrelated to either of the parent species (Chiba, 2005). Fossil materials used for the analysis were those previously described as Mandarina luhuana, which occurred in the Pleistocene and Holocene fissure, cave and dune deposits of Chichijima and Minamijima (Chiba, 1989, 1998). Four morphotypes (forma A–D) of M. luhuana were tentatively described in a previous taxonomic study (Chiba, 1989). Forma B occurred in the Pleistocene fissure and cave deposits and forma A in the fissure, dune and cave deposits of Chichijima and Minamijima. Forma C, initially described as Mandarina pallasiana, occurred in the dune deposits of Chichijima (Chiba, 1989) (Table 1, Figure 1). Forma D is not treated in the present study, because only one individual was found. Geological setting and 14C age of the fossil samples, except for location C8, were given by Chiba (1989 Chiba (1996). The age of the dune deposit of location C8 was newly estimated using the 14C dating method. Specimens used for the descriptions have been deposited in the University Museum, Tohoku University (TUMC), Institute of Boninology (CIBML), and the University Museum, University of Tokyo (UMUT).
Table 1.
List of samples used in the study.
Morphological analysis
The animals used for morphological analysis were all adults. Twelve measurements of the adult shells (Figure 2) were made with a Nikon model V16D profile projector with a digital micrometer (Appendix A). In total 9 variables, diameter (D), height (H), spire index (H/D), number of whorls (Wh), relative width of umbilicus (U/D), relative width of protoconch (Pw/D), peripheral angularity (Mw/Mh), peripheral height (Ph/Bh), index of suture depth (Sh/Sw), and ratio of thickness between outer lip and basal lip (Ol/Bl) were obtained from these measurements. The sample mean of each variable was compared statistically between different samples using ANOVA, with probabilities adjusted by Holm's method. Color patterns were classified following the coding method described by Chiba (1999b) (Appendix B). Difference in color patterns among samples was examined using extended Fisher's exact tests.
Figure 2.
(A) Shell measurements. A–I: shell height, F–N: diameter (D), D–H: height of body whorl (Bh), D–G: height of shoulder (Ph), F–S: width of margin of body whorl (Mw) (F–S is 1/10 length of F–G), Q–R: height of margin of body whorl (Mh) (Q and R are intersections between shell outline and a vertical line that passes point S), K–J: width of umbilicus (U), K–L: thickness of basal lip (Bl), P–O: thickness of outer lip (Ol), B–D: length of the next whorl (Sw), C–E: inflation of the next whorl (Sh).
(B) Protoconch and spiral lines on the shell. U–V: protoconch width (Pw), number of whorls from W to T: number of whorls of the shell (Wh). The position of the whorl on which the spiral lines disappear was scored as follows: 4, lines present after 0.5 whorl before the last whorl; 3, lines present after 1 whorl before the last whorl but absent on the 0.5 whorl before the last whorl; 2, lines present after 1.5 whorl before the last whorl but absent on the 1 whorl before the last whorl; 1, lines present after 2 whorls before the last whorl but absent on the 1.5 whorl before the last whorl. Shells without spiral lines were scored as 0.
Before calculation of the morphological distance, samples showing no difference in any of the variables were combined into a single sample. When the difference of the sample means was significant, the morphological distance between the samples was calculated for each variable by the formula 
, where Xi is the sample mean of the variable of the sample i, and vij is the average of the variance of the variable between the samples i and j. If no significant difference was found, a value of 0 was recorded for the morphological distance. Individuals in some populations of Mandarina possess a shell with a number of fine spiral lines on the surface. The lines often become obscure with growth and disappear before maturity. The position of the whorl on which the lines disappear (Li) was scored as shown in Figure 2. Because this score showed very small or no variation within samples, the distance of this variable between the samples was taken as the difference between the sample means of the score. Difference in shell color patterns between the samples was determined as Roger's distance (Rij) calculated from the frequency of each color morph as 
, where m is the number of color morphs and Cki is the frequency of color morph k in the sample i.
Discrimination model
An artificial neural network (ANN) is a system designed to learn from data in a manner emulating the learning pattern in the brain. ANN techniques are becoming standard classification and prediction tools for physicists, astronomers, engineers, computer and cognitive scientists, neurophysiologists, biologists, and philosophers because of their high discriminative power, less stringent statistical assumptions and because they have no limitation on the type and distribution of data (Ripley, 1996). Recent application to palaeontological data showed that the ANN approach provides more reliable estimation of palaeoclimate than traditional procedures (Malmgren and Nordlund, 1997; Malmgren et al., 2001). In taxonomy, the use of the ANN approach has increased recently in a significant manner, with most efforts centered on morphological classification of species (Culverhouse et al., 1996, Kim et al., 1997; Shinn et al., 2000; Baylac et al., 2003).
In the present study, a data matrix of the morphological distances among the extant samples was used to train the ANN for subsequent classification of fossil samples. Thus, the network learns the difference between the patterns of interspecific and intraspecific distances, and then applies that information to unknown samples.
The ANN used in the present study is feedforward with backpropagation training, and consists of an input layer, a single hidden layer and output layer (Figure 3). The input layer consists of 12 neurons, each of which receives data on one of the 12 shell variables. The input layer is connected to the hidden layer, with every input neuron connected to every hidden neuron. The hidden layer is connected to the output layer, which consists of two neurons, one for each possible state (same species or different species). The value at each hidden neuron consists of a linear weighted sum of all input neurons, with the weights allowed to change based on the training data. The output node values are the weighted sums of all the hidden node values. These connections used a logistic function, since a binary output (either same species or different species) is expected. The ANN learns by repeatedly passing through the data and adjusting its connection weights to minimize error (misclassification). The weights were adjusted iteratively until the outputs matched the true results. This process was repeated many times until the weights stabilized. The number of hidden neurons was allowed to change, and is optimized in this study to give the most accurate classification. Initial weights were selected randomly and the learning parameter (a weight that determined how rapidly the backpropagation process converged to a stable network) was also optimized. The classification accuracy was tested using the entire training set of data, and was indicated by root mean square error and a confusion matrix. In addition, to ascertain the learning ability of the network, the ANN was trained on a randomly selected 80% of the distance data set of the extant samples and tested on the remaining 20% of the samples. This process was repeated 100 times and the level of learning ability was evaluated. To compare with the classification ability of the ANN, linear discriminant analysis was also conducted using log-transformed data of the shell measurements. All statistical analyses and the linear discriminant analysis were done in R (R Development Core Team, 2005). The ANN analysis was done using the Weka system (Witten and Frank, 2005).
Figure 3.
Artificial neural network (ANN) for the analysis of shell morphological distance among extant samples. Each neuron is indicated by a circle. The values of 12 shell variables were passed into the neurons of the input layer, where they were fed into the hidden layer, scaled according to each neuron's individual weighting factor. The neurons of the hidden layer passed their weighted data to the two neurons of the output layer, which represent the result of classification, either of the same species or of different species. Each value in the neurons of the hidden and the output layers represents the sum of all weighted inputs. The number of the neurons of the hidden layer was optimized for application.
Results
Of the 39 extant samples, 35 were significantly different (P < 0.05) from all other samples in more than one characteristic. In total, 595 distance data sets from the extant samples were used to train the ANN. Initially, 2–20 hidden neurons were investigated and the optimum number yielding the lowest root mean square error (0.0045) in the training set was determined to be 7. Approximately 10,000 iterations with 7 hidden neurons were subsequently used to train the network.
All of the 595 distance data sets were correctly classified by the trained network. On the other hand, of these 595, 38 (6.4%) were incorrectly classified by the linear discriminant analysis. The trained ANN with 80% of the randomly selected distance data sets could classify the remaining 20% of the data sets with 0–3.4% inaccuracy, and with an average misclassification rate of 1.4%. Thus, the ANN trained with morphological data from the extant samples had enough classification accuracy to be appropriate for classification of the fossil samples.
There were no significant differences in any of the characteristics between the samples of forma A of M. luhuana in Minamijima (samples Lm1 and Lm2) nor between those in Chichijima (samples Lc2 and Lc7). The morphological distances among the fossil samples and among those and the extant samples were calculated after these morphologically indistinguishable samples were combined. The shells of the Minamijima samples of forma A differed significantly (ANCOVA, F = 91.7, P < 0.001) from the shells of the Chichijima samples of forma A in Wh and H/D (Figure 4). Shell sculpture of these samples was also distinct: the former possessed a number of fine spiral lines on the surface of the shell spire (mean Li = 1.07), whereas the latter possessed no such lines on the shell (Li = 0). In addition, the scores of shell color morph were clearly different between the samples from Minamijima and Chichijima (X2 = 20.8, p < 0.01): the former possessed shells with darker coloration in much higher frequency than the latter. Despite these distinctions between the shells of the Minamijima and Chichijima samples of forma A, the trained ANN classified them as belonging to the same species.
Figure 4.
Scatter plots of relative spire index (H/D) against number of whorls (Wh) in the Chichijima and Minamijima samples of Mandarina luhuana forma A.
The differences between forma A, forma B and forma C of M. luhuana lead to their classification as different species. These morphs of M. luhuana and the extant samples of Mandarina were also classified as different species. Hence, the three morphotypes of M. luhuana were discriminated as different species, and the two distinct populations of forma A were classified as different subspecies of forma A.
Discussion
By adopting the ANN discrimination model to consider morphological differences among extant species of Mandarina, the fossil samples previously thought to be a single species, M. luhuana, were discriminated into three species. High classification accuracy of the ANN model suggests that the morphological differences among the three fossil species are equivalent to the morphological differences among the extant species, which were discriminated based on reproductive isolation and molecular phylogenetic relationships. However, this does not necessarily imply that the definition of species in the fossil record is the same as for extant species. This is because the ANN classification model was based only on the association between overall patterns of differences in shell morphology and taxonomic status of the extant samples, and not necessarily on the differences in the specific characteristics associated with reproductive isolation or phylogenetic relationships. The biological meaning of the classification model developed by the ANN is not clear because of the highly nonlinear nature of the relationships among the variables in the model. It was difficult to estimate which variables reflect phylogenetic relationships and which reproductive isolation. Detailed investigation of the relationships among reproductive isolation, ecology, behavior and shell morphology is needed to resolve the above issues. Although the present classification method for fossil samples has these limitations, and further modeling using nonlinear, non-parametric statistical methods is needed for a better understanding of the mechanisms underlying these associations, the present procedure enabled as much biological information from the extant populations as possible to be included for classification of the fossil samples and alignment of fossil species as closely as possible with extant species.
Another limitation is that even if a method with a high classification power such as this ANN is used, it is essentially impossible to discriminate cryptic species. In Mandarina, fortunately, genetic studies have shown that there is no evidence of coexistence of different cryptic species that can be discriminated only by differences in molecular markers. Hence, this problem can be ignored in the present case.
Using rather “lumped” taxonomy, as in the present approach, tends to ignore information that would be important in understanding environmental change or in determining a particular span of geologic time. For example, a shift from one morphospecies to another through time provides a useful indication of environmental changes that have occurred, but if these morphospecies are lumped as a single species, we may lose a useful practical approach to evaluation of palaeoenvironments. To resolve this problem, the distinctive forms between which the differences are not equivalent to those between the extant species are treated as nominal subspecies. Although describing subspecies may not be recommended in certain taxonomic circumstances (Zink, 2004), it is necessary in the lumped taxonomy of fossils such as those in the present study. The Minamijima populations and Chichijima populations of forma A possess distinct shell morphologies that would be discriminated as different morphospecies if traditional approaches were applied. Hence, it is practical to describe these populations as nominal subspecies.
No intermediates were found between forma A and forma B, which therefore were easily classified as different species. However, no boundary can be drawn between successive forms along a phyletic lineage, and therefore, intermediates should be found between the two forms if the fossil sequence could be observed continuously under a finer time scale. However, the presence of intermediates does not affect the description of the species. For example, ring species, in which two reproductively isolated species are connected by a number of intermediate forms, are known in many extant taxa. The intermediate forms can be described as either of the two species or just as intermediate forms between the two species.
The present approach to discriminating fossil species cannot be applied to taxa of which all members have gone extinct. However, for groups of organisms including many extant species, the present method of classification provides a good approximation of the taxonomy of extant species to the fossil record.
Systematics
Order Stylommatophora
Family Bradybaenidae Pilsbry, 1934.
Genus Mandarina Pilsbry, 1894
Type species Mandarina mandarina (Sowerby, 1839)
Mandarina carinata sp. nov. Figures 5A, 5B, 5C.
Figure 5.
Shells of Mandarina carinata sp. nov. (A–C, holotype TUMC 25001), M. pallasiana (D–F, CIBML200701), M. luhuana minamijimana ssp. nov. (G–I, holotype TUMC 24996) and M. luhuana luhuana (J–L, TUMC 24998). Scale bar = 1 cm.
Mandarina luhuana (Sowerby, 1839) forma B. Chiba, 1989, p. 232, 233, 235, 237, fig. 9.2.
Type material.—Holotype: TUMC 25001 (Figures 5A, 5B, 5C), Minamizaki (loc. C2), Chichijima, Ogasawara. Paratypes: TUMC 25002 (n = 9), Minamizaki (loc. C2), Chichijima, Ogasawara; TUMC 25003 (n = 10), Minamizaki (loc. C2), Chichijima, Ogasawara; UMUT CM 18429 (n = 30), Minamizaki (loc. C2), Chichijima, Ogasawara.
Dimensions of holotype.—Diameter 47.2 mm; height 23.6 mm; 3 postembryonic whorls.
Shell morphology.—Shell very large and thick for the genus. Spire low and slightly domed. Suture very weakly impressed. Embryonic whorls 1.5–2.0; first whorl smooth, thereafter with a number of fine spiral lines on the surface. Surface of postembryonic shell with a number of fine spiral lines and fine radial growth lines. Shell uniformly brownish in color. Body whorl with sharp peripheral angle. Position of the periphery on the uppermost part of the body whorl. Body whorl depressed above the periphery and inflated below the periphery. Base of shell rather flat. Umbilicus distinctively open. Aperture oblique, slightly compressed above periphery and inflated below the periphery, with slightly reflected outer margin and broadly reflected basal margin. Columellar and basal lip thickened. Slightly thickened parietal callus extending slightly to left of columella in basal view. Whorls of postembryonic shell 2.9–3.3 in number.
Distribution.—Minamizaki, Chichijima, Ogasawara (latest Pleistocene).
Remarks.—This species was described as a variant of Mandarina luhuana (Sowerby, 1839) by Chiba (1989). However, it differs from M. luhuana in having a larger shell with higher spire, narrower umbilicus and thinner apertural lip. In addition, M. luhuana does not have a strong peripheral angle and many spiral lines on the body whorl. M. carinata differs from Mandarina pallasiana (Pfeiffer, 1850) in having a larger shell, higher position of the periphery on the body whorl, vertically longer aperture and more inflated lower body whorl of the shell. In addition, the color pattern of M. carinata (uniformly brown) is less than 30% frequency in M. pallasiana.
Mandarina pallasiana (Pfeiffer) Figures 5D, 5E, 5F.
Helix pallasiana Pfeiffer, 1850, p. 67.
Helix pallasiana Pfeiffer. Tryon, 1886, p. 131, pl. 44, figs. 49, 50.
Mandarina pallasiana (Pfeiffer). Kuroda, 1930, p. 206, pl. 14, fig. 2.
Mandarina pallasiana (Pfeiffer). Habe, 1969, p. 19, 23, pl. 1, figs. 1, 2.
Mandarina pallasiana (Pfeiffer). Habe, 1973, p. 51, pl. 4, fig. 4.
Mandarina pallasiana (Pfeiffer). Minato, 1978, p. 44, 50, pl. 4, figs. 9, 10.
Mandarina luhuana forma C (Pfeiffer). Chiba, 1989, p. 232, 233, 235, 237, fig. 9.3.
Materials.—CIBML200701 (n = 1), Kitafukurozawa (loc. C8) Chichijima, Ogasawara; CIBML200702 (n = 9), Kitafukurozawa (loc. C8) Chichijima, Ogasawara; UMUT CM18430 (n = 4), Kitafukurozawa (loc. C8) Chichijima, Ogasawara.
Shell morphology.—Shell large, slightly thick for the genus. Spire low. Suture very weakly impressed. Embryonic whorls 1.5–2.0; first whorl smooth, thereafter with a number of fine spiral lines on the surface. Surface of postembryonic shell with a number of fine spiral lines and fine radial growth lines. Shell white, always with 2–4 brown bands, or uniformly brown without any bands. Body whorl with sharp peripheral angle. Periphery close to the middle of the body whorl. Rather straight outline below periphery of body whorl. Base of shell rather convex. Umbilicus distinctly open. Aperture oblique, slightly compressed above and below periphery, with slightly reflected outer and basal margins. Columellar and basal lip thickened. Slightly thickened parietal callus extending slightly to left of columella in basal view. Whorls of postembryonic shell 2.7–3.2 in number.
Distribution.—Kominato and Kitafukurozawa, Chichijima, Ogasawara (Holocene).
Mandarina luhuana minamijimana subsp. nov. Figures 5G, 5H, 5I.
Mandarina luhuana (Sowerby, 1839). Habe, 1973, p. 51, p. l4, figs. 1, 2, 3.
Mandarina luhuana (Sowerby). Minato, 1978, p. 43, 50, pl. 4, figs. 7, 8.
Mandarina luhuana (Sowerby) forma A. Chiba, 1989, p. 232, 233, 235, 237, fig. 9.1.
Type material.—Holotype: TUMC 24996, Minamijima (loc. M1), Ogasawara. Paratypes: TUMC 24997 (n = 9), Minamijima (loc. M2), Ogasawara; UMUT CM 18428 (n = 30), Minamijima (loc. M1), Ogasawara.
Dimensions of holotype.—Diameter 40.3 mm; height 21.5 mm; 3 whorls.
Shell morphology.—Shell large, very thick for the genus. Spire very low and domed. Suture strongly impressed. Embryonic whorls 1.5–2.0; first whorl smooth, thereafter with a number of fine spiral lines on the surface. Surface of postembryonic shell shiny with a number of rough radial growth lines. A number of fine spiral lines on the surface of the postembryonic shell until one whorl before the adult aperture. Shell yellow, always with 2–4 brown bands, or uniformly brown without any bands. Body whorl with no or very weak peripheral angle. Umbilicus very wide and distinctly open. Body whorl inflated at the lower part. Base of shell rather flat. Aperture oblique, slightly compressed above and below periphery, with broadly reflected outer and basal margins. Thickened parietal callus extending slightly to left of columella in basal view. Whorls of postembryonic shell 2.8–3.5 in number.
Distribution.—Minamijima, Ogasawara (Holocene).
Remarks.—This subspecies differs from Mandarina luhuana luhuana in having a higher spire, a smaller number of whorls, a relatively narrower umbilicus and a thinner shell. On the surface of the spire, this subspecies possesses fine spiral lines that are absent in M. l. luhuana. In addition, shell color is darker than M. l. luhuana. This subspecies includes individuals with relatively very high spire, but this form belongs only tentatively to this subspecies. Further examination using larger samples is needed to determine the taxonomic status of this high-spired form.
Mandarina luhuana luhuana (Sowerby, 1839) Figures 5J, 5K, 5L.
Helix luhuana Sowerby, 1839, p. 143, p. 35, fig. 4.
Nanina ruschenbergeri Pilsbry, 1890, p. 186, text-fig. 3.
Mandarina ruschenbergeri (Pilsbry). Pilsbry, 1902, p. 141.
Mandarina luhuana (Sowerby). Kuroda, 1930, p. 206, pl. 14, figs. 3, 4.
Materials.—TUMC 24998 (n = 1), Minamizaki (loc. C7), Chichijima, Ogasawara; TUMC 25000 (n = 15), Minamizaki (loc. C7), Chichijima, Ogasawara; UMUT CM18427 (n = 30), Minamizaki (loc. C2), Chichijima, Ogasawara.
Shell morphology.—Shell large, very thick for the genus. Spire very low. Suture strongly impressed. Embryonic whorls 1.5–2.0; first whorl smooth, thereafter with a number of very fine spiral lines on the surface for the subsequent 0.5–1 whorl. Surface of postembryonic shell shiny with a number of rough radial growth lines. Shell color white, always with 2–4 brown bands. Body whorl with no or very weak peripheral angle. Umbilicus very wide and distinctly open. Body whorl inflated at the lower part. Base of shell rather flat. Aperture oblique, slightly compressed above and below periphery, with broadly reflected outer and basal margins. Thickened parietal callus extending slightly to left of columella in basal view. Whorls of postembryonic shell 3–3.6 in number.
Distribution.—Minamizaki, Chichijima, Ogasawara (Holocene).
Acknowledgments
I express my sincere thanks to I. Hayami, K. Tanabe, A. Davison, the South Kanto branch of the Ministry of the Environment, the Ogasawara branch office of the Tokyo Metropolitan Government and the Institute of Boninology for assistance in this survey, and to K. Tomiyama and I. Okochi for providing materials. Special thanks are due to R.H. Cowie and Y. Kano for providing valuable and helpful comments on this manuscript. This study was conducted under permits from the Agency for Cultural Affairs and the Ministry of the Environment. This research was supported by the Global Environment Research Fund (F-051).
