ENDODONTOID LAND SNAILS FROM PACIFIC ISLANDS (Mollusca: Pulmonata: Sigmurethra) : Family Endodontidae ./. . ". ' i Museum of Natural History Chicago, Illinois LIBRARY OF THE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN AUG 1 6 1977 ENDODONTOID LAND SNAILS FROM PACIFIC ISLANDS (Mollusca: Pulmonata: Sigmurethra) Part I Family Endodontidae Alan Solem Curator of Invertebrates Field Museum of Natural History With the technical assistance of Barbara K. Solem Field Museum of Natural History Chicago, Illinois Published by Field Museum of Natural History October 29, 1976 This monograph was prepared with the support of National Science Foundation grant No. DEB75-14048. Any opinions, findings, conclusions, or recommendations expressed in such a publication are those of the author and do not necessarily reflect the views of NSF. Library of Congress Catalog Card Number: 76-9516 US ISSN 0015-0754 INTKI) IN THE UNITED STATES OF AMERICA II v. L CONTENTS LIST OF FIGURES VI LIST OF TABLES X INTRODUCTION 1 Acknowledgements 5 PREVIOUS STUDIES 7 MATERIAL STUDIED 9 METHODS OF ANALYSIS n Variation in adult shells 11 Nature of quantitative data presented 12 Sample bias 13 Measurement reliability 14 Criteria for species recognition 15 Nature of comparative remarks 17 PATTERNS OF MORPHOLOGICAL VARIATION 19 Size and shape variations 19 Body whorl contour 21 Spire protrusion 24 Umbilical contour 26 Brood-chamber formation 27 Effects of whorl increment 30 Shell sculpture 30 Types and growth patterns 30 Family-level differences 36 Other sculptural elements 37 Patterns of variation 39 Correlated variations 44 Rib spacing 44 Rib reduction 47 Functional significance of sculpture 50 Other external shell features 50 Apertural barriers 52 Parietal barriers 54 Columellar barriers 57 Palatal barriers 59 Parietal and palatal barrier traces and expansion correlations 63 Barrier growth 63 Microdenticulations 65 Barrier numbers and length 65 Degree of apertural narrowing 71 Summary of barrier variation 72 Gross anatomy 72 Genital system 73 Ovotestis 73 Hermaphroditic duct 75 Talon and carrefour 76 Albumen gland 77 Prostate and uterus 77 Terminal male genitalia 78 Terminal female genitalia 83 III IV Pallial complex 84 87 Digestive system c Free muscle system "4 Nervous system 94 External body features 94 Patterns of elongation "4 Summary of anatomical variation 98 HABITAT RANGE AND EXTINCTION 100 PHYLOGENY AND CLASSIFICATION 102 Phylogenetic position of the endodontoid snails 102 Family classification of the endodontoids 105 Phylogeny within the Endodontidae 107 Portrait of a generalized endodontid 109 Identifiable major trends 109 Phylogenetic conclusions HO Fossil endodontoid land snails 116 Previous generic classifications 118 Proposed generic classification 119 SYSTEMATIC REVIEW 121 List of the taxa 122 Geographic keys to the genera 124 Genus Minidonta, new genus 126 Key to the genus 130 Group of Minidonta micra 130 Group of Minidonta micraconica 135 Group of Minidonta rotellina 139 Group of Minidonta anatonuana 139 Group of Minidonta simulata 146 Genus Mautodontha, new genus 151 Key to the genus 154 Subgenus Mautodontha, s. s 154 Subgenus Garrettoconcha, new subgenus 162 Genus Anceyodonta, new genus 178 Key to the genus 186 Genus Cookeconcha, new genus 207 Genus Kleokyphus, new genus 224 Genus Opanara, new genus 227 Key to the genus 232 Genus Rhysoconcha, new genus 255 Genus Ruatara, new genus 265 Genus Orangia, new genus ....276 Key to the genus ....278 Genus Australdonta, new genus ,...289 Key to the genus ....294 Genus Taipidon, new genus ....314 Key to the genus ....317 Group of Taipidon analogica ....318 Group of Taipidon centadentata ....330 Group of Taipidon varidentata ....333 Genus Planudonta, new genus 335 Key to the genus 337 Genus Rikitea, new genus 342 Genus Nesodiscus Thiele, 1931 345 Key to the genus 35^ Genus Nesophila Pilsbry, 1893 365 Genus Kondoconcha, new genus 36g Genus Endodonta Albers, 1850 371 Genus Pseudolibera, new genus 383 Genus Libera Garrett, 1881 385 Key to the genus 390 Genus Gambiodonta, new genus 431 Key to the genus 434 Genus Thaumatodon Pilsbry, 1893 444 Key to the genus 448 Group of Thaumatodon decemplicata 448 Group of Thaumatodon hystricelloides 453 Group of Thaumatodon subdaedalea 461 Genus Priceconcha Solem, 1973 465 Genus Zyzzyxdonta, new genus 466 Genus Adadonta, new genus 467 Key to the genus 473 ZOOGEOGRAPHY 488 SUMMARY 493 REFERENCES 494 APPENDIX: Anatomical terms 501 INDICES..., 502 LIST OF FIGURES 1. Frequency distribution of whorl counts in adult Libera fratercula 12 2. Frequency distribution of shell height in adult Libera fratercula 12 3. Frequency distribution of Height /Diameter ratio in adult Libera fratercula 13 4. Frequency distribution of shell diameter in adult Libera fratercula 13 5. Method of measuring specimens 14 6. Frequency distributions of shell height in the Endodontidae 20 7. Frequency distributions of shell diameter in the Endodontidae 20 8. Frequency distributions of Height /Diameter ratios in the Endodontidae 20 9. Frequency distributions of whorl counts in the Endodontidae 20 10. Frequency distributions of D/U ratios in the Endodontidae 21 11. Pattern of umbilical size in Rapa and Mangareva Island Endodontidae 22 12. Frequency distributions of rib counts in the Endodontidae 22 13. Pattern of rib spacing on the body whorl 22 14. Effects of changes in peripheral whorl contour 23 15. Effects of changes in spire protrusion 25 16. Patterns of phyletic change in umbilical contours 27 17. Mean shell height in species with and without a brood chamber 28 18. Mean shell diameter in species with and without a brood chamber 29 19. Mean Height/Diameter ratio in species with and without a brood chamber 29 20. Mean whorl count in species with and without a brood chamber 29 21. Sculpture of Rhysoconcha atanuiensis and Thaumatodon decemplicata (Mousson) 31 22. Sculpture in Gambiodonta 32 23. Postnuclear sculpture of Australdonta raivavaeana 33 24. Patterns of radial surface sculpture in selected Endodontidae 34 25. Sculpture of Minidonta hendersoni 35 26. Postapical sculpture of Cookeconcha decussatulus (Pease) 36 27. Sculpture of Mautodontha (M.) aoraiensis 37 28. Sculpture of Aaadonta constricta constricta (Semper) 38 29. Sculptural details on Aaadonta constricta constricta 39 30. Apical microsculpture in Cookeconcha 40 31. Apical and postnuclear sculpture of Libera bursatella bursatella (Gould) and Nesodiscus taneae (Garrett) 41 32. Method of sculptural measurements in the Endodontidae 42 33. Correlation between rib count and shell diameter in Opanara areaensis areaensis from Station 383...4S 34. Correlation of ribs and ribs/mm, in Opanara areaensis 43 35. Relationship of mean shell diameter to major rib spacing in Minidonta, Anceyodonta, and Gambiodonta 48 36. Apertural barrier terminology and numbering system 52 37. Apertural barrier form in the Endodontidae 52 38. Variation in parietal barriers 53 39. Shell sculpture and denticle structure in Opanara areaensis 55 40. Palatal barrier resorption surface in Thaumatodon spirrhymatum 64 41. Palatal barrier sculpture in Thaumatodon hystricelloides (Mousson) 64 42. Palatal barrier sculpture in Hawaiian Endodontidae 66 43. Method of measuring parietal and palatal barrier lengths 70 44. Ovotestis structure and position in the Endodontidae and Pacific Island Charopidae 74 45. Hermaphroditic duct variation in Endodontidae 75 46. Vas deferens entrance and penial retractor muscle insertion patterns 78 47. Pilaster cross-sectional patterns in typical Endodontidae 80 48. Spermathecal insertion patterns 84 VI VII 49. Anatomy of Taipidon petricola decora 85 50. Pallial cavity length variations 86 51. Radular teeth of Libera fratercula rarotongensis 88 52. Radular teeth of Taipidon petricola decora 89 53. Radular teeth of Endodonta fricki (Pfeiffer) 90 54. Radular teeth of Thaumatodon hystricelloides (Mousson) 91 55. Animal length and shell whorl count in elongated taxa 95 56. Penis length and shell diameter in the Endodontidae 96 57. Phyletic diagram of the Endodontidae 110 58. Levels of specialization in the Endodontidae, size range within genera, and hypothesized directions of evolutionary change Ill 59. Computer generated phylogeny of Thaumatodon, Zyzzyxdonta, and Aaadonta 114 60. Computer generated phylogeny of Taipidon and generalized Rapan genera 115 61. Computer generated phylogeny of Minidonta, Mautodontha, Kleokyphus, and Cookeconcha subpacificus 117 62. Minidonta manuaensis, M. inexpectans, and M. rotellina 131 63. Minidonta micra and M. hendersoni 133 64. Anatomy of Minidonta, Mautodontha, Rhysoconcha, and Ruatara 136 65. Minidonta micraconica and M. gravacosta 138 66. Correlation of height and diameter in Minidonta anatonuana, M. haplaenopla, and M. planulata 140 67. Correlation of diameter and D/U ratio in Minidonta anatonuana, M. haplaenopla, and M. planulata 141 68. Minidonta anatonuana and M. sulcata 142 69. Minidonta planulata and M. haplaenopla 144 70. Minidonta simulata and M. taunensis 147 71. Minidonta taravensis and M. extraria 149 72. Mautodontha (M.) boraborensis and M. (M.) ceuthma 152 73. Mautodontha (M.) zebrina and M. (M.) daedalea 157 74. Mautodontha (M.) zimmermani and M. (M.) aoraiensis 160 75. Mautodontha (Garrettoconcha) saintjohni and M. (G.) consobrina 167 76. Mautodontha (Garrettoconcha) maupiensis, M. (G.) punctiperforata, and M. (G.) imperforata 169 77. Mautodontha (Garrettoconcha) parvidens, M. (G.) subtilis, and M. (G.) rarotongensis 172 78. Mautodontha (Garrettoconcha) consimilis, M. (G.) acuticosta, and M. (G.) unilamellata 175 79. Relationship of height to diameter in Mautodontha consimilis and M. acuticosta 176 80. Relationship of whorls to diameter in Mautodontha consimilis and M. acuticosta 177 81. Anceyodonta ganhutuensis and A. subconica 187 82. Anceyodonta constricta, A. alternata, and A. andersoni 190 83. Anceyodonta difficilis and A. soror 193 84. Relationship of height to diameter in Anceyodonta soror and A. difficilis 194 85. Relationship of H/D ratio to D/U ratio in Anceyodonta soror and A. difficilis 195 86. Anceyodonta sexlamellata 197 87. Anceyodonta densicostata and A. labiosa 200 88. Anceyodonta obesa 201 89. Anceyodonta obesa 202 90. Anceyodonta hamyana 205 91. Relationship between mean diameter and mean ribs/mm, in Cookeconcha 211 92. Cookeconcha subpacificus 212 93. Cookeconcha stelluhis 218 94. Cookeconcha thaanumi and C. luctiferus 219 95. Kleokyphus callimus and K. hypsus 225 96. Anatomy of Opanara depasoapicata, O. bitridentata, O. duplicidentata, and O. a. areaensis 228 97. Anatomy of Opanara altiapica, O. m. megomphala, O. m. tepiahuensis, O. fosbergi, and O. perahuensis 229 98. Relationship of height to diameter in Opanara bitridentata, O. caliculata, O. depasoapicata, and O. duplicidentata 230 99. Distribution of Opanara in the Mt. Perahu region 231 100. Distribution of Opanara megomphala, O. caliculata, and O. altiapica 232 101. Distribution of Opanara areaensis 232 102. Opanara bitridentata and O. duplicidentata 234 VIII 103. Carrefour region in Opanara duplicidentata 238 104. Opanara areaensis and subspecies 243 105. Opanara caliculata and O. altiapica 247 106. Opanara m. megomphala and O. m. tepiahuensis 250 107. Opanara per ahuensis, O. fosbergi, and O. depasoapicata 252 108. Structure of parietal barriers in Rhysoconcha variumbilicata and Kondoconcha othnius 255 109. Distribution of Rhysoconcha 258 110. Ribs and rib spacing in Rhysoconcha species and hybrids 259 111. Size and shape frequencies in Rhysoconcha species and hybrids 260 112. Rhysoconcha variumbilicata and R. atanuiensis 263 113. Ruatara koarana and R. o. oparica 266 114. Ruatara oparica normalis and R. o. reductidenta 268 115. Barrier number variation in Ruatara oparica 270 116. Distribution of Ruatara oparica 273 117. Distribution of Orangia 277 118. Umbilical closure in Orangia c. cookei 277 119. Proportions of closed and barely perforate umbilici in Orangia 278 120. Ribs and ribs/mm, in Orangia 279 121. Anatomy of Orangia 280 122. Correlation of height and diameter in Orangia cookei 281 123. Shells of Orangia species and subspecies 282 124. Microsculpture of Australdonta raivavaeana 290 125. Anatomy of Australdonta 293 126. Apertural barrier numbers in Australdonta species 294 127. Australdonta magnasulcata and A. pseudplanulata 295 128. Australdonta degagei and A. rimatarana 297 129. Australdonta tapina and A. yoshii 301 130. H/D and D/U ratio correlation in Australdonta degagei, A. rimatarana, and A. tapina 302 131. Relationship of height and diameter in Australdonta degagei, A. rimatarana, and A. tapina 303 132. Australdonta r. radiella and A. r. rurutuensis 305 133. Australdonta raivavaeana and A. tubuaiana 308 134. Frequency distributions of apertural barriers in populations of Australdonta raivavaeana 309 135. Correlation of parietal and columellar barrier numbers in Australdonta raivavaeana 310 136. Relationship of height and diameter in Australdonta raivavaeana and A. tubuaiana 312 137. Australdonta pharcata and A. ectopia 313 138. Anatomy of Taipidon petricola, T. fragila, and T. varidentata 320 139. Anatomy of Taipidon semimarsupialis and T. centadentata 321 140. Taipidon p. petricola and T. p. decora 322 141. Taipidon woapoensis and T. octolamellata 325 142. Taipidon anceyana and T. marquesana 327 143. Taipidon analogica and T. semimarsupialis 329 144. Taipidon centadentata 332 145. Taipidon fragila and T. varidentata 334 146. Origin of umbilical sculptural pattern in Planudonta 336 147. Anatomy of Planudonta 338 148. Planudonta subplanula 340 149. Planudonta concava, P. intermedia, and P. matauuna 341 150. Rikitea insolens ...344 151. Umbilical mucus cover in Nesodiscus taneae 345 152. Shape variation in Nesodiscus taneae 348 153. Early whorl measurement standard for Nesodiscus 349 154. Anatomy of Nesodiscus f ictus 350 155. Nesodiscus huaheinensis and N. taneae 353 156. Nesodiscus obolus form obolus and N. o. form acetabulum 355 157. Relationship of height and diameter in Nesodiscus obolus ...356 158. Nesodiscus obolus form celsus ...357 159. Nesodiscus cretaceus and N. fabrefactus \ar.piceus ...359 160. Nesodiscus fabrefactus and N. f ictus ...361 161. Nesodiscus magnificus ...365 IX 162. Kondoconcha othnius 369 163. Anatomy of Endodonta fricki 372 164. Anatomy of Endodonta fricki 373 165. Anatomy of Endodonta lamellosa, Nesophila tiara, Cookeconcha jugosus, and C. hystricellus 374 166. Endodonta ekahanuiensis 375 167. Hawaiian Endodonta 379 168. Pseudolibera lillianae 384 169. Apical sculpture of Libera b. bursatella 387 170. Body whorl deflection and umbilical closure in Libera recedens 389 171. Anatomy of Libera b. bursatella and L. micrasoma 396 172. Anatomy of Libera cookeana and L. fratercula rarotongensis 399 173. Libera cookeana and L. micrasoma 401 174. Libera bursatella bursatella and L. b. orofenensis 403 175. Libera gregaria and L. recedens 404 176. Libera dubiosa 406 177. Libera spuria and L. garrettiana 408 178. Libera umbilicata and L. retunsa 411 179. Libera streptaxon and L. jacquinoti 414 180. Umbilical exit from brood chamber in Libera streptaxon 415 181. Libera incognata and L. heynemanni 416 182. Libera f. fratercula and L. f. rarotongensis 422 183. Ribs and ribs/mm, correlation in Libera fratercula 424 184. Libera (L.) subcavernula and L. (L.) tumuloides 427 185. Umbilical closure in Gambiodonta pilsbryi aukenensis 432 186. Gambiodonta agakauitaiana and G. mangarevana 435 187. Gambiodonta p. pilsbryi and G. p. aukenensis 437 188. Gambiodonta tumida and G. mirabilis 439 189. Gambiodonta grandis 442 190. Distribution of Thaumatodon and Zyzzyxdonta 447 191. Anatomy of Thaumatodon hystricelloides and T. decemplicata 449 192. Thaumatodon multilamellata 450 193. Thaumatodon decemplicata and T. laddi 452 194. Thaumatodon euaensis and T. hystricelloides 457 195. Anatomy of Thaumatodon euaensis 459 196. Thaumatodon vavauensis and T. corrugata 460 197. Thaumatodon subdaedalea and T. hystricelloides 462 198. Zyzzyxdonta alata 466 199. Anatomy of Aaadonta (c.) constricta and A. (f. ) fuscozonata 471 200. Anatomy of Aaadonta kinlochi 472 201. Size and shape variation in Aaadonta kinlochi, A. constricta, and A. irregularis 473 202. Proportionate differences between A. angaurana and A. c. constricta 473 203. Aaadonta c. constricta and A. irregularis 475 204. Aaadonta constricta komakanensis, A. c. babelthuapi, and A. angaurana 477 205. Proportionate differences between Aaadonta pelewana and A. fuscozonata 479 206. Aaadonta f. fuscozonata and A. f. depressa 480 207. Aaadonta pelewana 482 208. Aaadonta kinlochi and apertures of A. c. constricta and Thaumatodon hystricelloides 486 LIST OF TABLES I. Pacific Island endodontoid taxa 2 II. Hawaiian endodontoids 3 III. Dates of species descriptions 3 IV. Summary of material studied 9 V. Frequency distribution of specimens examined per species level taxon 10 VI. Size and shape variation in relict, allopatric Rapan Opanara 17 VII. Whorl contour and whorl count correlations 24 VIII. Spire protrusion and whorl count correlation 24 IX. Size and shape correlations with umbilical contour changes 26 X. Umbilical contour and spire protrusion 28 XI. Whorl count correlated size increment 30 XII. Rib counts in Opanara areaensis from Station 383 42 XIII. Pattern of rib spacing in Opanara areaensis from Station 383 43 XIV. Correlation of microradial counts with major rib counts and shell diameter in the Endodontidae 44 XV. Shell diameter and rib spacing in the Endodontidae 44 XVI. Correlation between rib spacing and shell diameter in the Endodontidae 45 XVII. Size and degree of sculpture reduction in larger Endodontidae 46 XVIII. Shell size and sculpture reduction in Polynesian Endodontidae 47 XIX. Shell size and subperipheral rib reduction in larger Endodontidae 49 XX. Body whorl descension and shell size 50 XXI. Body whorl contour and rate of descent 51 XXII. Correlation between sulci and body whorl contour 51 XXIII. Presence of trace barriers in the Endodontidae 54 XXIV. Correlation between number of parietal barriers and descension of 2nd parietal barrier 56 XXV. Correlation between anterior descension of 2nd parietal barrier, diameter, and whorl count in species with 2 parietal barriers 56 XXVI. Correlation between diameter and superior structure of parietal barriers 56 XXVII. Loss of columellar barrier 58 XXVIII. Correlation between columellar barrier height and positional relation to lip edge 58 XXIX. Columellar barrier position and recession 59 XXX. Shell size and columellar barrier position 59 XXXI. Shell size and columellar barrier height 59 XXXII. Palatal barrier shape and length correlations 60 XXXIII. Palatal barrier recession and size correlations 60 XXXIV. Phyletic correlations of palatal barrier recession 61 XXXV. Size correlation with palatal barrier expansion 61 XXXVI. Patterns of barrier expansion 62 XXXVII. Reduction and loss of palatal barriers 62 XXXVIII. Correlation of unusual palatal barrier expansion with parietal barrier expansion 63 XXXIX. Percentage distribution of parietal barrier numbers 65 XL. Percentage distribution of palatal barrier numbers 67 XLI. Phyletic correlation of parietal barrier numbers 68 XLII. Size correlation of parietal barrier numbers 69 XLIII. Phyletic correlations of palatal barrier numbers 69 XLIV. Size correlation of palatal barrier numbers 69 XLV. Correlation of parietal and palatal barrier numbers 70 XLVI. Parietal barrier length and size correlations 70 XLVII. Palatal barrier length and size correlations 70 XLVIII. Degree of apertural narrowing by barriers ...71 X XI XLIX. Correlation of apertural narrowing and body whorl contour 72 L. Phyletic representation of dissected taxa 73 LI. Correlation of penial pilaster shape and relative size 81 LH. Sympatric variability in penis size and pilaster patterns in Rapan taxa 82 LIII. Sympatric variability in penis size and pilaster patterns in Tahitian species 82 LIV. Sympatric variability in penis size and pilaster patterns in Marquesan species 82 LV. Radular tooth size and numbers in Nesophila, Cookeconcha, Taipidon, and Planudonta 93 LVI. Shell size and jaw structure in the Endodontidae 94 LVII. Patterns of aulacopod radular dentition 97 LVIII. Major differences between Endodontidae and Charopidae 97 LIX. Minor differences between Endodontidae and Charopidae 98 LX. Shell parameters of the Endodontidae 109 LXI. Shell parameters for levels of organization 113 LXII. Frequency distribution of species/genus in Pacific Island endodontoids 120 LXIII. Range of variation in Minidonta 129 LXIV. Local variation in Minidonta 145 LXV. Range of variation in.Mautodontha (Mautodontha) 155 LXVI. Range of variation in Mautodontha (Garrettoconcha) 163 LXVII. Local variation in Mautodontha 164 LXVIII. Character states typifying A nceyodon ta 180 LXIX. Range of variation in Anceyodonta 181 LXX. Local variation in Anceyodonta 182 LXXI. Geographic distribution and relative abundance of Mangarevan Endodontidae 185 LXXII. Percentage species composition of Mangarevan Anceyodonta in selected samples 186 LXXIII. Barrier variation in Anceyodonta hamyana 207 LXXIV. Local variation in Cookeconcha 208 LXXV. Range of variation in Opanara and Kleokyphus 223 LXXVI. Local variation in Opanara, Ruatara koarana, and Kondoconcha othnius 236 LXXVII. Ribs/mm, in Opanara 237 LXXVIII. Parietal barrier variation in Opanara bitridentata 237 LXXIX. Local variation in Opanara areaensis 240 LXXX. Rib variation in Opanara areaensis 241 LXXXI. Range of variation in Rhysoconcha, Ruatara, Kondoconcha, and Orangia .257 LXXXII. Differences between Rhysoconcha variumbilicata and R. atanuiensis 259 LXXXIII. Local variation in Rhysoconcha atanuiensis, R. variumbilicata, and hybrid populations 261 LXXXIV. Rib variation in Rhysoconcha variumbilicata, R. atanuiensis, and hybrid populations 262 LXXXV. Frequency distribution of ribs/mm, in Ruatara oparica 269 LXXXVI. Rib variation in Ruatara oparica 271 LXXXVII. Local variation in Ruatara oparica 272 LXXXVIII. Rib variation in Orangia 283 LXXXIX. Local variation in Orangia 284 XC. Range of variation in Australdonta 291 XCI. Local variation in Australdonta 299 XCII. Range of variation in Taipidon 316 XCIII. Local variation in Taipidon 319 XCIV. Rib spacing and whorl diameter in Planudonta 337 XCV. Range of variation in Planudonta, Rikitea, and Nesodiscus 343 XCVI. Local variation in Nesodiscus 346 XCVII. Early whorl size in Nesodiscus 349 XCVIII. Local variation in Endodonta 370 XCIX. Ribs and rib spacing in Libera 388 C. Range of variation in Libera 392 CI. Early whorl diameter in Libera 393 CII. Local variation in Mangarevan Expedition Libera 395 CIII. Local variation in Society Island Libera 409 CIV. Island variation in Libera fratercula 420 CV. Local variation in Libera fratercula 421 CVI. Local variation in Libera subcavernula, L. tumuloides, and L.jacquinoti 429 CVII. Range of variation in Gambiodonta 433 XII CVIII Local variation in Gambiodonta. 443 CIX Range of variation in Thaumatodon and Zyzzyxdonta 445 ex. Local variation in Thaumatodon and Zyzzyxdonta 454 CXI Range of variation in Aaadonta 469 CXII. Percentage of adults in Aaadonta 470 CXIII. Distribution of Aaadonta 472 CXIV. Local variation in Aaadonta... ....485 INTRODUCTION The small, radially ribbed, usually flammulated pulmonate land snails that dominate, in terms of species numbers, the faunas of southern temperate regions traditionally have been lumped into a single family unit, the Endodontidae. Several family units are involved and many genera cannot yet be assigned to a particular family. In discussing various genera and structures, I have found it convenient to use several terms in a restricted fashion. These are: charopid member of the family Charopidae charopinine structure agreeing with the typical Charopidae pattern endodontid member of the family Endodontidae endodontinine structure agreeing with the typical Endodontidae pattern endodontoid having the general aspect of several families punctid member of the family Punctidae punctinine structure agreeing with the typical Punctidae pattern Other terms either conform to standard malacological usage or are defined in the text. This is the first of two monographs on the endodontoid land snails of Polynesia, Micronesia, and Fiji. It reviews the larger and older of the family groups, the Endodontidae. A second contribution 1 will cover the specifically less diverse and more recent Punctidae and Charopidae, together with detailed zoogeographic and faunistic analyses. Together these papers are based on more than 26,000 specimens and review 285 species-level taxa in 45 genera. Most of these were previously unreported, with 84 per cent of the genera and 54 per cent of the species described as new (table I). This survey, started in 1960, is incomplete, since about 290 unnamed Hawaiian species-level taxa are preserved in the collections of the Bernice P. Bishop Museum (table II). The latter represent a monophylet- ic radiation involving limited supraspecific, but exten- sive specific level differentiation. Analysis of the approximately 58,000 Hawaiian specimens would have required far more time than was available. Hence consideration of the Hawaiian fauna has been restrict- ed to a brief synopsis of the 30 previously named taxa, one new description, and sufficient dissections to place these groups within the Endodontidae. 'Hereafter referred to as Part II. In addition, collecting efforts on the Pacific Islands are far from being comprehensive. A few hours in 1970 on Thithia and Tuvutha Islands in the Lau Archipelago of Fiji resulted in finding two new taxa, Priceconcha tuvuthaensis and Thaumatodon spirrhymatum (Solem, 1973d), that are only cross- referenced in this study. Many islands in Lau have not been collected on at all, the higher elevations of the Society and Marquesan Islands have been in- adequately sampled, and relatively fragmentary collec- tions have been made on some islands of the Palau Group. I suspect that these 573 species-level taxa of endodontoid snails may represent only 75 per cent of the fauna extant in 1900. Unfortunately, a high percentage is now extinct through habitat alterations by man, predation by accidentally introduced ants, and possibly because of competition from introduced snails. These figures show that the endodontoid snails are the most diverse land snail group found in the Pacific, substantially surpassing in numbers the fa- belled Hawaiian Amastridae (294 species and sub- species, Zimmerman, 1948, pp. 99-100) and mainly Hawaiian Achatinellidae (ca. 200 species, Cooke and Kondo, 1960). One other family-level grouping has an equally wide distribution and shows high generic level diver- sity. The zonitoid families Helicarionidae and Zoni- tidae (H. B. Baker, 1938b, 1940, 1941) total 32 genera with 266 Pacific Island species (H. B. Baker, 1941, p. 347). Compared with the 45 genera and 573 species level endodontoid taxa, this is proportionately greater generic differentiation and lower specific differ- entiation. For the zonitoid taxa an average of 8.31 species/genus compares with an average of 13.2 species/genus in the endodontoid families. This un- doubtedly reflects the fact that the zonitoids are much more recent colonizers of the Pacific and represent a greater number of colonizing stocks. There has been less time for local differentiation. Seemingly there has been a greater niche balance in terms of colonizing stocks and hence less opportunity for diversification. Other families that are significant in terms of species numbers on Pacific Islands include the Succi- neidae, Pupillidae (sensu lato), Diplommatinidae, He- licinidae, and Assimineidae. Probably each has about 150 species on the Pacific Islands. Generally, each family shows only one or two major centers of specific diversity: the Succineidae in Tahiti and Hawaii, SOLEM: ENDODONTOID LAND SNAILS TABLE I. - PACIFIC ISLAND ENDODONTOID TAXA Endodontidae Punctidae & Charopidae Subtotals Genera Known New 5 19 19 Species Known New 81 88 i i i *"-! Subspecies Known New 2 lU 38 126 131 21 TOTALS 257 23 I. Three species listed as new but not described are omitted. Pupillidae in Hawaii, Assimineidae in the Mariana and Caroline Islands, Diplommatinidae in Micronesia and Fiji. The great diversity found in the endodontoid taxa has not been recognized previously because the group has been virtually ignored during this century. Some 76 per cent of the described species were named prior to 1890 (table III). Outside of a few radular sketches published in the 1870's to 1890's, and partial dis- sections of three Endodonta from Hawaii (Cooke, 1928), no anatomical information has been recorded in the literature. The reasons for this are simple. The species are very small (96.4 per cent are less than 7 mm. in maximum size), secretive inhabitants of litter or may be found on moss-covered tree trunks in dense and undisturbed forests. They are found only by the most skilled collectors. Endodontoid snails are of no known economic importance and are sparsely repre- sented in even the largest museum collections of the world. My own interest in these taxa was sparked by the recovery of Miocene to Pleistocene fossil endodontoid snails from the deep-core drillings on Bikini, Eniwetok, and Funafuti Atolls (Ladd, 1958, 1968). I had worked on endodontoid species from the New Hebrides (Solem, 1959a) and examined the limited Polynesian material in mainland American museums. Seeing the fossil species suggested that a revision of the living endodontoid taxa from Polynesia and Micronesia might provide insights into the historical zoogeography of that region, present a time dimension to the colonization of the islands, and permit some observa- tions concerning the rate of evolution in these island populations. I was led to believe that perhaps 100 living species might be involved in such a survey. Exposure to the vast collection resources of the Bernice P. Bishop Museum soon wrought drastic revisions in this program. The fossils barely show species or species group differentiation from living taxa, add only one major modification to a current geographic range, and provide no significant data concerning possible evolutionary rates (pp. 116-118). The huge quantity of material, potentially 84,000 specimens of 573 species level taxa, soon led to quite a different focus than simple species sorting and class- ification formation. Although the endodontoid snails do not show determinate growth upon reaching sexual maturity, there are distinct patterns of growth change (pp. 11- 12) that enable separation of adult from juvenile examples. The many large sets and essentially synchronic samples in the Bishop Museum suggested that detailed meristic analysis of populational vari- ation should replace the conchologically traditional, non-statistical means of studying variability and in delineating species. Most endodontoid snails have a highly complex surface sculpture, show moderate shape variation, and have a few to many complex barriers within the aperture of the shell. Thus a wealth of characters was available for study. Most of this basic species level analysis was done during the period in which phenetics enjoyed its youthful flush of enthusiasm. Since the previous classification of the Pacific Island endodontoid snails dated from a checklist (Pilsbry, 1893a) that used such dichotomies as "teeth- no teeth" plus "rounded shell-angled shell" for generic placement, the temptation to experiment with phyletic clustering techniques could not be resisted. By this stage of the study (mid-1960's), dissections had indicated that at least two major groups were represented in the fauna. These groups differed strikingly in their anatomy, but showed equally striking similarities in shell appearance. Several shells INTRODUCTION TABLE II. - HAWAIIAN ENDODONTOIDS Previously described Genus En do do nt a Cookeconcha Nesophila 9 IT k In Bishop Museum Collection Number of Number of Number of species subspecies catalogued sets Punctum 77-79 105 12 5-9 55 55 9 339 170 TOTALS 31 199-205 119 5,197 from widely separated archipelagoes were virtually identical in overall appearance. Even microscopic shell features appeared the same under 32-96 X mag- nification, but the animals had totally different anatomical structures. The computer programs avail- able to me at that time did not allow use of characters for which information was lacking in regard to certain taxa. Hence only conchological variables and no anatomical data could be utilized. Beautiful clusters of conchological convergences from disparate areas (figs. 59-61) were produced by these phenetic attempts. While this numerical dabbling did not produce a usable classification and added few phyletic insights, it did serve the extremely important functions of greatly increasing the number of shell characters observed, requiring me to systematically record the variational state of each character for each species, to focus on the patterns of change for various characters, and to start systematically reviewing the effects of particular character variations on the standard measurable shell features. A more important feedback came in correlating shell variations as analyzed and recorded during this phase of the study with anatomical variations dis- covered during the dissections. Initial focus in the dissecting was to sort out major groups, then to study patterns probably indicative of generic clusters, and finally to study the variations shown within a genus, particularly in regard to sympatric congeners. A second and much more critical phase of the anatomical survey involved detailed analysis of the variational TABLE III. - DATES OF SPECIES DESCRIPTIONS Charopidae Endodontidae Hawaiian non-Hawaiian 1820's 2 1830's 1 181+0 's b 5 1850's 3 3 1860's 5 8 1870's 1 16 1880's 3 18 1890 's 3 1900's 8 1910's 1920's 1930's 1 19*10 's 1950 's 2 . I960 's 6 6 5 1 It 3 2 TOTALS 30 53 4 SOLEM: ENDODONTOID LAND SNAILS patterns within organ complexes, and finally attempts to determine probable directions of character change in the various systematic units. Perhaps a majority of the species was extinct before this project started, and only 100 of 283 taxa could be dissected. With the high degree of conchological convergence noted above, a major problem involved how to classify species whose anatomy could not be studied. Fortunately, the advent of scanning electron microscopy provided a tool that enabled finding highly obscure but stable shell features correlating with the major anatomical types (pp. 30-37, 63-65). Considerable time has been spent in dissecting endodontoid taxa from Norfolk Island, Lord Howe Island, New Caledonia, New Zealand, South America, and South Africa in an attempt to determine: 1) the names that should be applied to the Pacific Island family units; 2) how closely the Charopidae are related to the Endodontidae and if any transitional species are still extant; and 3) what are the basic anatomical patterns of variation in the Charopidae. While the name problem is settled, at the present time the other two problems are still unresolved. Phyletic interpretations are received today with greater enthusiasm when outgroup comparisons are made with more primitive and presumably ancestral taxa. The problems of circularity in reasoning are very great when work is limited to discussion of trends within the group or groups being reviewed. Attempts to pinpoint taxa that might be ancestral to the Endodontidae have met with a total lack of success to date. Dissections of Succineidae, which many authors recently have stated to be a primitive land snail, suggested instead (Solem, 1969b, In press B) that this is an advanced group belonging to the same suborder as the endodontoid taxa, while dissections of several orthurethran and mesurethran taxa demonstrated that these have nothing in common with the anatomical patterns of the endodontoid taxa and cannot be considered ancestral to them (Solem, unpublished). At present I can only state that the Endodontidae have the greatest number of primitive features found in any sigmurethran land snail. I cannot point to any land snail family as being a possible ancestor to the Endodontidae (pp. 102-104). Hence meaningful out- group comparisons have not been possible. What began as a zoogeographically oriented faunistic survey changed drastically in orientation. Analysis of the patterns in both conchological and anatomical variations, scanning electron microscope studies of shell and radular structure (Solem, 1972a, 1972c, 1973a, 1973b), revision of the family- level classification for endodontoid snails (pp. 105-107), and most of the new ideas concerning slug evolution plus a revised ordinal-level classification of the stylom- matophoran pulmonates (Solem, In press B) have developed from these endodontoid revisions. The above discussion outlines the goals and evolutionary chronology of this study. Obviously many aspects pursued involved use of many more characters than are needed to enable keying out species or characterizing genera. Several of these objectives are discussed in subsequent papers or in the second monograph. But documentation of these conclusions requires that the basic data be readily available. Computer programs are far more sophisticated than those used in the earlier stages of this study and, in particular, factor analysis could be expected to aid the interpretation of this data. Neither computer time nor personal computer competence is available for pur- suing these aspects. Equally important is the matter of species limits in island organisms. For those accustomed to working with continental faunas, the bewildering pattern of diversity and the, at times, minute conchological differences between species come as a surprise. Fewer than 15 per cent of the previously described species could be identified with one of the delineated population complexes without direct comparison with type or syntype specimens. Descriptions and illustra- tions that focused on differentiating only a few among many species were totally inadequate when the percentage of total taxa available for study increased. From many islands, such as Tahiti and Moorea, only partial and preliminary single, high altitude collecting transects have been made up to the present time. Sampling of additional populations can be expected to yield not only undescribed taxa, but populations that will cause revision in some of the species concepts outlined below, and add much additional data on microgeographic variation. Detailed descriptions and illustrations will greatly aid work on future collections. Finally, a good proportion of the material reported was collected from reasonably restricted stations that are localized with as much precision as the lack of man-made markers and inadequate maps permitted. It will be possible, except where extinction has taken place, to sample the same populations on an allochron- ic basis. By recording the available data on variation in local populations, as well as summary data on species units, suggestions both as to the locations for new field collecting and the interpretation of the results from such field collections will be facilitated. Hence the diagnoses, descriptions, and discussions of taxa include far more data than is necessary to separate related species at our current level of species recognition and understanding of species limits. When- ever a character or character complex was found to show systematically significant variation between any pair of species units, the state of this character was recorded for each species in the family. For obvious reasons, this has been limited to optically visible conchological characters and such anatomical data as was available for particular species. All data on which statements concerning variational trends and charac- INTRODUCTION ter correlations are made thus have been recorded in the systematic section and are readily available for others to use in further analyses. Particular emphasis in this volume is laid in analyzing the variations in shell features and their correlations with size and shape variations. Equivalent patterns are found in the Charopidae, subject to alterations correlated with shifts in ecology. Discussion of the basic adaptive value of these variations is deferred until the second volume to permit inclusion of the full data. Following a brief summary of previous reports, material studied, and methods of analysis, the patterns of morphological and ecological variation are reviewed. Discussions of family level classification, probable phylogeny and generic classification precede geograph- ic keys and the main systematic review. A short, preliminary zoogeographic analysis concludes this report. ACKNOWLEDGEMENTS Success of this project has been possible only because of work done previously by the late C. Montague Cooke, Jr. and Yoshio Kondo. For 46 years (Kondo and Clench, 1952) Dr. Cooke assiduously amassed unparalleled reference collections of Pacific Island non-marine mollusks at the Bernice P. Bishop Museum, Honolulu. His own large collections from the Hawaiian Islands resulted from 10 personal ex- peditions. Much additional material resulted from the activities of colleagues and amateurs inspired by his example. During the 1930's, Cooke was responsible for three expeditions to Polynesia (Mangarevan Ex- pedition, 1934), Micronesia (Micronesian Expedition, 1935-1936), and Fiji (Henry G. Lapham Expedition, 1938). These trips provided the bulk of the material upon which this report is based. The wealth of preserved soft parts and shells resulting from these efforts stands as a permanent monument. Cooke never rushed into print with fragmentary discoveries. His list of publications is comparatively short. From 1939 to 1941 he spent considerable time sorting the endodontid material into species, making a few preliminary notes, and overseeing preparation of illustrations by Yoshio Kondo. For a limited number of species, he drew up preliminary descriptions of shell features. Cooke preferred to let a project sit for several years and thus benefit from more mature reflection. He also was more interested in the Tornatellinidae. These factors combined to shelve the endodontid work in 1941. The magnificent revision of the Achatinellidae and the Tornatellinidae published by Cooke and Kondo (1960) resulted from Cooke's endeavors between 1941 and his death in 1948. Without Cooke's efforts, this material would not have been collected. Without his years of museum work, it would not have been available for study. Without his preliminary sorting and notes, my task would have required at least two additional years. Cooke's preliminary work had been restricted to shell features. Descriptions had been written only for Henderson Island, Mangareva, and the Tuamotu species. In reviewing material from these areas, I agreed with Cooke's species limits in most cases, made minor changes in others, and completely altered some species. While I take full responsibility for errors, I wish to recognize Cooke's great contributions not only through the taxa named after him, but by having the species he wrote up in a preliminary fashion cited subsequently as described by "Cooke and Solem," or "Solem and Cooke" in cases where I altered species limits. All published diagnoses and descriptions were prepared by myself, but where so indicated, species authorship should be cited as a joint responsibility. Yoshio Kondo began collecting for Dr. Cooke during the Mangarevan Expedition while he was a crew member on the Islander. After the voyage he served as assistant, then as collaborator and finally became successor to Dr. Cooke. Kondo prepared illustrations for most of the Mangareva, Marquesas, Society, and Rapa Island species from 1939-1941. It was at his urging and encouragement that this monograph was started. During our work in Honolulu from August through December 1961, Dr. Kondo provided every facility and aid possible. Through his cooperation it was possible to borrow both soft parts and shell material for more detailed analysis and to publish his prepared illustrations. Figures prepared by Yoshio Kondo are indicated in the figure explanations by "YK." I am deeply grateful for his help and encouragement during all stages of this project. The aid of the Bernice P. Bishop Museum in providing working space, allowing study of materials, and permitting publication of the many shell drawings has been crucial. I am greatly in their debt. Establishing the identity of previously described species and reviewing the characteristics of extralimital species referred to Pacific Island taxa occupied an inordinate amount of time and effort. For assistance in locating material and/or loan of specimens, I am indebted to the following curators and institutions, which are listed in chronological order of visits and /or borrowings: Dr. A. W. B. Powell, Auckland Institute and Museum, Auckland; Dr. R. K. Dell, National Museum of Natural History, Wellington; Dr. Charles Fleming, New Zealand Geological Survey, Lower Hutt; Dr. D. F. McMichael, formerly of the Australian Museum, Sydney; Mrs. Hope Black and Dr. Brian Smith, National Museum of Victoria, Melbourne; the late Charles Gabriel, Melbourne; the late Bernard C. Cotton, South Australian Museum, Adelaide; Dr. A. Zilch, Natur-Museum Senckenberg, Frankfurt-a.-M.; Dr. E. Fischer-Piette, Museum National d'Histoire Naturelle, Paris; Dr. A. Magne, Museum d'Histoire Naturelle, Bordeaux; Dona Emilia Garcia San Nicolas, 6 SOLEM: ENDODONTOID LAND SNAILS Museo Nacional de Ciencias Naturales, Madrid; Dr. E. Tortonese, Museo Civico di Storia Naturale, Genoa; Dr. Eugene Binder, Musee de la Ville, Geneva; Dr. Lothar Forcart, Naturhistorisches Museum, Basel; Dr. H. Burla and Dr. H. Jungen, Zoologisches Museum der Universitat, Zurich; Dr. J. Knudsen, Universitetets Zoologiske Museum, Copenhagen; Dr. Bengt Huben- dick, Naturhistoriska Museet, Goteborg; Dr. Charlotte Holmquist, Naturhistoriska Riksmuseet, Stockholm; Mrs. W. S. S. van Benthem Jutting van der Feen, Dr. S. van der Spoel, Dr. H. Coomans, Zoologisch Museum, Amsterdam; Dr. C. O. van Regteren Altena, Rijksmu- seum van Natuurlijke Historie, Leiden; Dr. W. Adam, Institut Royal des Sciences Naturelles de Belgique, Brussels; Mr. Norman Tebble, Mr. John Peake, and Mr. S. Peter Dance, British Museum (Natural His- tory), London; Dr. William J. Clench, Museum of Comparative Zoology, Harvard; Dr. W. K. Emerson, American Museum of Natural History, New York; Dr. R. Tucker Abbott and Dr. Robert Robertson, Acad- emy of Natural Sciences, Philadelphia; Dr. Harald A. Rehder and Dr. Joseph Rosewater, National Museum of Natural History, Washington; Dr. Juan Jose Parodiz, Carnegie Museum, Pittsburgh; Dr. Vincent Conde, Redpath Museum, Montreal; Mr. Allyn Smith, California Academy of Sciences, San Francisco; Dr. A. Rodger Waterston, Royal Scottish Museum, Edinburgh; Mr. Colin Matheson, National Museum of Wales, Cardiff; and Dr. A. Myra Keen, Stanford University. While many of the institutions cited above do not have material listed in this particular report, their collections and materials have been indispensible to its completion. For assistance during field work in Tahiti, Samoa, Fiji, New Caledonia, New Zealand, and Australia, I am indebted in particular to Mr. Laurie Price of Kaitaia, New Zealand, who collected on Rarotonga, Samoa, Lord Howe Island, Norfolk Island, Lau Archipelago, Viti Levu, New Caledonia, New Hebrides, and Tonga in connection with this work; Dr. Pierre Cassiau, Papeete; Mr. L. Devambez, formerly of New Caledonia and Fiji; Mr. Michael Watt, Apia; Dr. A. W. B. Powell; and Dr. D. F. McMichael. I have been extremely fortunate in having the services of several excellent illustrators. The work of Miss Margaret Anne Moran (MM), Mr. Sander Heilig (SH), Miss Marcia Oddi (MO), Mr. Samuel H. Grove (SG), Miss Carole Wrigley (CW), Mrs. Claire Kryczka (CK), and Miss Patricia Rill (PR) is gratefully acknowledged. Together with the large nucleus of drawings prepared by Yoshio Kondo (YK) for the late C. M. Cooke, Jr., their efforts have enabled presenta- tion of phylogenetic and morphological trend data in an understandable form, and provided identification aids far superior to optical photographs. Without their contributions, the value of this studv would have been greatly diminished. In particular, Miss Moran's assis- tance has been invaluable. Mounting and labelling of the figures, preparation of maps and graphs, and mounting of the tabular material has been handled quite competently by Mrs. Jane Calvin, Mrs. Dorothy Karall, Mrs. Claire Kryczka, and Miss Marian Pahl. Scanning electron microscope photographs have been made at the American Dental Association, Chicago and Alpha Research and Development Company, Blue Island, Illinois. I am deeply indebted to Dr. Harvey Lyon, Mr. John Lenke, Mr. George Najarian, and Mr. John Brown for assistance and advice. Photographic work by Mr. Ferdinand Huysmans and Mr. John Bayalis has been invaluable. Initial statistical analysis of the material mea- sured during 1961 and 1962 was done by Mrs. Barbara Solem and Mrs. Robin Napier. Subsequent data processing by Mrs. Sandra Rendleman, Mrs. Rita Mecko, and myself was required in preparation of tabular material and study of material obtained more recently. Most tabular material was collected into final form and typed by Mrs. Rendleman and Ms. Jayne Freshour. Manuscript typing and proofreading by Mrs. Rita Mecko, Mrs. Sandra Rendleman, Mrs. Dorothy Karall, Mrs. Lynda Hanke, Mrs. Alice Burke, Mrs. Nancy Kozlowski, Mrs. Paula Steele, Ms. Barbara Walden, Ms. Jayne Freshour, Mrs. Sharon Bacoyanis and the following students from Antioch and Wilson Colleges Victoria Leuba, Carl Sainten, Kam B. Louis, and Jeanne Sinderman is gratefully acknowledged. Considerable technical assistance in mounting jaws and radulae was provided by Mrs. Pamela Hall, Ms. Barbara Walden, and Mrs. Lynda Hanke. For patience in serving as a sounding board for ideas and for many helpful suggestions during various stages of this study, I wish to thank Dr. Yoshio Kondo, the late Dr. H. B. Baker, the late Dr. Fritz Haas, Dr. J. Felsenstein, and Mr. Henry Dybas. Grateful acknowledgement is made to the Nation- al Science Foundation whose generous support through grants G-16419, GB-3384, and GB-6779 has been instrumental in illustration preparation plus needed museum study and the field work involved. Receipt of Grant No. DEB 75-14048 from the National Science Foundation enabled publication of this monograph. The patience of Field Museum of Natural History in permitting such a lengthy time investment and exploration of so many side facets of knowledge in the course of almost 12 years has been critical, while their support in 1961-1962 of Barbara Solem through a Dee Fellowship made the data gathering possible. For expert and patient help with proofreading and indexing, I am deeply indebted to Mrs. Sharon Bacoyanis. PREVIOUS STUDIES The only summary studies of Pacific Island endodontoid land snails are the check list and descriptions of Pease (1871a), the collection of illustra- tions and brief diagnoses assembled by Tryon (1887), and the critical checklist and revised classification presented by Pilsbry (1893-1895). Tryon's classification was highly artificial and conservative. Pilsbry 's effort was peripheral to his focus on the morphology of helicoid taxa and phyletic revision of the higher land snails. His treatment of the endodontoid taxa obviously was hastily done, but contained several significant advances in classification. It has served as the basis from which the generic lists presented by Thiele (1931) and Zilch (1959-1960) were derived. After 1900, only the descriptions of Pilsbry and Vanatta (1905, 1906), the nomenclatural catalog of Libera by Ponsonby (1910), an extremely significant anatomical survey of some Hawaiian species by Cooke (1928), and the reports on fossil endodontoid taxa by Ladd (1958, 1968), and Ladd et al. (1970) require special mention. Other reports are simple compilations of names in faunistic catalogues (Gude, 1913; Caum, 1928; Germain, 1932), scattered descriptions of one or two species in faunal reports (Clapp, 1923; Cockerell, 1933; I. Rensch, 1937; Dell, 1955; Solem, 1960), or reports on faunistic collections without descriptions (Cooke, 1934; Aubert de la Rue and Soyer, 1958). The available knowledge is essentially pre-1900, with all the deficiencies for modern systematic research that this implies. Fortunately, two of the workers, Andrew Garrett and Albert Mousson, were far ahead of their time in terms of systematic concepts and recognizing the need for precise geographic documentation. Garrett was an American missionary stationed on a number of different South Pacific Islands from the 1860's until his death in 1887. At first he sent specimens to W. Harper Pease, a merchant in Hawaii, who described a total of 17 species (Pease, 1861, 1864, 1866, 1867, 1868, 1870, 1871a), unfortunate- ly, often with scant regard for locality data (Garrett, 1881, p. 390). Later Garrett started to do his own describing, first in a pair of strictly descriptive papers (Garrett, 1872, 1874), then in faunistic surveys of Rurutu, Austral Islands (Garrett, 1879), the Cook Islands (Garrett, 1881), Society Islands (Garrett, 1884), Samoan Islands (Garrett, 1887b), Fiji (Garrett, 1887a), and the Marquesan Islands (Garrett, 1887c). Including eight species collected by Garrett but described by W. Harper Pease, Andrew Garrett was responsible for 40 of the 128 previously known taxa. Not only were his descriptions detailed and accurate to the limits of optical viewing, but his extensive field experience was reflected in his precise localization of species and many observations concerning the degree of sympatry or allopatry on the same island. Most zoologists ignored such data until well into this century. Garrett's data has been summarized below, but consultation of his papers remains mandatory for anybody attempting field work in the South Pacific. Garrett's views of speciation were quite modern and few of his conclusions based on field work have been modified by subsequent studies. He was located far from the museum centers of Europe and North America, so that inevitable errors in synonymy occured, but these do not lessen the substance of his contributions. The only major limitation that can be placed on his work concerns the apparent restriction of his collecting efforts to reasonably low elevations. High-altitude collections in the Marquesas by the members of the Bishop Museum "Pacific Entomological Survey" (Adamson, 1935, 1936), in the Society Islands by members of the Mangarevan Expedition from the B. P. Bishop Museum (Cooke, 1935, p. 51), and on Rar- otonga by a Field Museum Expedition (Solem, 1972b) have discovered many large and conspicuous species unknown to Garrett and failed to find a high percentage of his species. This suggests quite strongly that his efforts were restricted to the lower forested areas, whose fauna is now extinct, and that he did not attempt to reach higher elevations in his collecting. Albert Mousson never visited the South Seas. His material was obtained from the Hamburg trading firm Godeffroy, mostly as a result of the efforts by Eduard Graeffe. The eight species described by Mousson (1865, 1869, 1870, 1871, 1873) are carefully delineated, well illustrated and accompanied by detailed locality data. His faunal accounts of species from Fiji, the Ellice Island, Tonga, and Samoa are inferior to Garrett's only because field observations are lacking. The faunal report on the Caroline Islands by Otto von Mollendorff (1900) and the paper on Guam species by J. Quadras and Mollendorff (1894) are quite competent, as are the many Philippine studies that included the description of one endodontid (Mollendorff, 1888). The 11 species named in these studies are almost equaled by the 10 described by C. F. Ancey (1889a, b, c, 1899, 1904). Most of these are from Hawaii, and are briefly mentioned below. The two 8 SOLEM: ENDODONTOID LAND SNAILS Libera from the Society Islands and the single species from Mangareva were recognizable from his descrip- tions. The summary of Hawaiian land snails by Ancey (1889d) formed the basis for the more familiar compilations of Baldwin (1893), Sykes (1900), and Caum (1928), although it is infrequently cited today. From a wide variety of sources and localities, L. Pfeiffer (1845, 1846a, b, 1850a, 1853a, 1856, 1858, 1859a, 1862) named nine species. Otherwise there are the scattered descriptions of Ferussac (1824, 1840), Anton (1839), Hombron and Jacquinot (1841), Gould (1844, 1846a), Mighels (1845), Reeve (1851-1854), Cox (1870), Semper (1874), Liardet (1876), Tapparone-Canefri (1883), Tryon (1887), Beddome (1889), Sykes (1896), E. A. Smith (1897), Pilsbry and Vanatta (1905, 1906), Clapp (1923), Cockerell (1933), Dell (1955), and Solem (1959b, 1960). To summarize the above data, Table III indicates the dates of description for the species recognized as valid in this study. It is obvious that the major contribution occurred between 1860 and 1890, when 57 per cent of the species were named. This is the primary period of activity for Pease, Garrett, Mousson, and Ancey, who account for 52 per cent of the previously named species. Since 1910, only 10 per cent (13 species) have been named, and all of these are fossil or extralimital to the main area of study. As indicated above, essentially no faunistic collecting was done except by the Bishop Museum program from the time of Mollendorff to the start of this study. Not only little faunal data accumulated, but there was practically no data recorded concerning the anatomy of endodontoid taxa from the Pacific Islands. Semper (1874, pp. 135-136, pi. 16, fig. 18) figured the central radular tooth of a Tahitian specimen identified as Libera bursatella and briefly discussed the cusp structure. W. G. Binney (1875, p. 248, pi. 21, fig. 6) illustrated several teeth and recorded the radular formula of the Rarotongan Libera tumuloides (Gar- rett) as 10-7-1-7-10. Subsequently, W. G. Binney (1885, pp. 88-89, pi. 2, figs. L-N) figured and briefly discussed the radulae of Libera tumuloides (Garrett), Nesodiscus huaheinensis (Pfeiffer) (formula 12-6-1-6-12), and Mautodontha parvidens (Pease) (formula 7-4-1-4-7) as "Endodonta incerta Mousson," a nude name under which this species has been widely distributed in older collections. Pilsbry (1893-1895, p. 23, pi. 9, fig. 34) illustrated the radular teeth of Libera recedens Garrett. No information concerning other anatomical features were recorded until the work of Cooke (1928) on three species of Endodonta from Oahu, Hawaiian Islands. Unfortunately, this excellent paper has been overlooked or misinterpreted by subsequent authors. Considerable information has accumulated concerning the external anatomy and radular struc- ture of Australian and New Zealand taxa, primarily through the efforts of Charles Hedley (1891, 1893a) and Henry Suter (1890, 1891a, b, 1892a, b, c, 1893a, b, c, d, 1894a, b, c, d, 1901, 1903, 1913). Recently Frank Climo (1969a, b, 1970, 1971a, b) has published several revisions of New Zealand taxa utilizing genital anatomy in addition to shell and radular features. His work presents a large quantity of useful data and important observations, but as outlined below (pp. 106-108), we have very different interpretations of character weighting and phylogeny. Prior to this study a high percentage of the species was unknown, and virtually no anatomical in- vestigations, no study of interpopulational variation, and only limited ecological information were available. As documented on pp. 118-119, classification was still based on convenient "either-or" pigeon holes. Hence not only has it been necessary to do much work that is already accomplished for better-known groups, but the hunt for criteria to use in species recognition and clustering has been quite time consuming. The lack of previous work also necessitates considerable prelimi- nary discussions concerning the characters used and their validity as "key" criteria. MATERIAL STUDIED As summarized in Table IV, some 26,000 speci- mens were studied in detail. Perhaps an additional 6,000 specimens, mainly involving sets with mixed species and localities that originated from the W. Harper Pease collection, were examined quickly, but neither measured nor listed. Pease himself was careless in handling his collection, noting in a letter to a correspondent that his small daughter delighted in playing with shells in the cabinets (Alison Kay, pers. comm.). In addition, many years later during shipment of Pease's collection from England to Harvard Univer- sity, apparently there was extensive mixing of sets when cabinets were tilted and handled. In subsequent years, these shells have been traded widely to other museums and amateur collectors. Virtually all traded Pease material that I examined contained more than one species, often living on different islands. While for the non-Hawaiian Endodontidae, the material examined per species-level taxon averaged 120.3, for the Charopidae it was only 75.9 speci- mens/species-level taxon. It is difficult to decide exactly how much of this is caused by artifacts of collecting and how much results from a true difference in relative abundance. The charopid taxa include a number of species from Melanesia and Indonesia, areas in which there has been far less intensive land snail collecting activity. Charopids also include more species from islands such as Niue, Rotuma, and various of the Ellice Islands where collections were made by non- malacologists and hence lower quantities of materials obtained. These factors probably reduced the average number of specimens by about 10. On the other hand, in the outer Polynesian Islands the Endodontidae were perhaps the dominant ground stratum snail. Where still extant, they can be collected in quantity from limited areas. In contrast, the charopids from the big islands of Fiji and Samoa, for example, are far less TABLE IV. - SUMMARY OF MATERIAL STUDIED Endodontidae Hawaiian non-Hawaiian Punctidae Charopidae TOTAL Species Level Taxa 31 2 96 283 Specimens Examined 238 18,530 1 7,288 26,063 abundant than many helicarionids and rarely have been collected in any quantity. On some of the Micronesian Islands apparently the charopids locally are as abundant as many of the Polynesian Endodon- tidae. Probably in the Melanesian-Fiji area the charopids are not abundant under any circumstances. Certainly the charopids with apertural barriers of Fiji, Tonga, Samoa, and neighboring islands could only be classified as rare. Punctids are a marginal group in the area of study. Hence the differing material per species numbers reflect both differential abundance and bias in collec- ting. Table V summarizes the abundance data on a slightly more refined basis. The rarity aspect of some charopid taxa shows very clearly in the increased number of species known from 1-3 specimens and decreased 4-8 specimen grouping. Otherwise the per- centage distributions are quite comparable. Data concerning species abundance on particular islands will be presented in the zoogeographic analysis accompanying Part II of this monograph. With the exception of pre-1900 species that have not been collected subsequently, the listing of speci- mens examined is confined to adequately localized and measured materials. Throughout the text the following abbreviations indicate the repository of the specimens: AMS Australian Museum, Sydney ANSP Academy of Natural Sciences, Philadelphia BMNH British Museum (Natural History), London BPBM Bernice P. Bishop Museum, Honolulu Brussels Institut Royal des Sciences Naturelles de Belgique, Brussels Cardiff National Museum of Wales, Cardiff FMNH Field Museum of Natural History, Chicago MCZ Museum of Comparative Zoology, Harvard University Paris Museum National d'Histoire Naturelle, Paris RSM Royal Scottish Museum, Edinburgh SMF Natur-Museum Senckenberg, Frankfurt USNM National Museum of Natural History, Washington Zurich Zoologisches Institut der Universitat, Zurich Most of the material studied was in the Bernice P. Bishop Museum. In addition to samples of Garrett's and Ancey's materials from the last century, Cooke had assiduously accumulated materials from various Pacific Islands. E. H. Bryan and Y. Kondo collected extensively on Guam, P. H. Buck from some of the Cook Islands, Cooke and Wray Harris in American Samoa, and Harry Ladd brought back highly sig- nificant collections from some of the Lau Archipelago in Fiji. But it was the three expeditions sponsored by Bishop Museum in the 1930's that were most 10 SOLEM: ENDODONTOID LAND SNAILS TABLE V. - FREQUENCY DISTRIBUTION OF SPECIMENS EXAMINED PER SPECIES LEVEL TAXON Number and Percent of Species Endodontidae Punctidae & Charopidae Number of Hawaiian non-Hawaiian Specimens 06- 1-3 l*-8 9-19 21-60 61-100 101-260 300-882 1,000-1,900 TOTALS 6 22(lU. 3%) 10 29(18.8%) 7 2l*(15.6%) 2 33(21.1*%) 13(8.1*%) 19(12.3%) 9(5.8%) 5(3-2%) 31 15 1 * 22(22.1*%) 7(7.1%) 13(13.3%) 26(26.5%) 8(8.1%) 17(17.3%) !*(!*.!%) 1(1.0%) productive. The Mangarevan Expedition from April 15 through October 28 in 1934 (Cooke, 1935), the Micronesian Expedition from December 8, 1935 to June 10, 1936 (Gregory, 1936, p. 40), and the Henry G. Lapham Expedition to Fiji from June 27 through September 28, 1938 (Buck, 1939, pp. 29-30) provided 70 per cent of the material on which this monograph is based. The intensity of collecting can be indicated by the statement of Cooke in Gregory (1936, p. 15) that 37,593 land snails were collected on Rapa and 30,695 on Mangareva during the Mangarevan Expedition. Much of the field work on Rapa was accomplished under abominable weather conditions (Zimmerman, 1938, pp. 3-4), and on only a portion of the 31-day stay was "full day" effort possible. The magnitude of these collections is awesome. The focus of malacological collecting was "On the trail of the Tornatellinidae" (Kondo and Clench, 1952, p. 17). The endodontoid snails were a very secondary consideration. Despite this the 4,078 Rapan endodontids are 10.8 per cent of the material collected, while the 2,274 endodontids from Mangareva are 7.4 per cent of the total shells. Most of the material was collected within a four- year period. The same people, C. M. Cooke, Y. Kondo, and E. Zimmerman, gathered the bulk of these collections. The collectors, collecting techniques, and field procedures were the same, so that comparisons of abundance and local distribution between archi- pelagoes are possible. Curatorial methods at the Bishop Museum differ from those used elsewhere and require some ex- planation. As is customary, new field collections from single stations are sorted into species. Whereas most museums would assign a different catalogue number to each set, i.e., all material of one species collected at one locality at one time, each lot of the Bishop Museum mollusks is sorted in several growth stages adult, paraneanic, metaneanic, ananeanic, and embryo (see Pilsbry and Cooke, 1914-1916, pp. x, xi for stage definitions). Each stage is given a separate, but usually consecutive, catalogue number. Sometimes gerontic individuals, unusual color or shape variations are segregated under yet other numbers. Thus what would be given a single number in most other museums, may have as many as 10 different numbers in the Bishop Museum collection. Soft parts are extracted from each shell by means of a water jet (Kondo and Clench, 1952, pp. 27-28) and material from one set is stored in tiny homeopathic vials. The latter are grouped into pint jars and shelved in numerical sequence. The dried shells are housed in small pill boxes, generally 1 1 A inch in diameter, with up to 48 such pill boxes placed in a covered cardboard tray. The latter are stored in open stacks. Only the catalogue number and a very abbreviated locality are with the specimens. Continual reference to the catalogue and field note books is necessary to retrieve full locality and ecological information. This is a highly efficient and effective system of housing and storing quantities of minute material, keeping dry and alcohol storage separate, and enabling age-class recognition of anatomical material. With species under the size of perhaps 5 mm. this system causes little difficulty. With larger shells, the limita- tions of space within individual pill boxes tend to produce bias in sorting. Invariably the gerontic and larger adult individuals are clustered separately from the adults that are closer to average in size. "Large" and "average" adults have separate catalogue num- bers. Measurements of such species as Nesodiscus fictus, many Endodonta, Libera bursatella bursatella, and the larger Gambiodonta had to take this sorting bias into account. In total some 80 per cent of the specimens cited are from the Bishop Museum collections. Perhaps half of the remaining is either from the Field Museum collections or was obtained by myself and Mr. Laurie Price of Kaitaia, New Zealand for this project. My own collecting has been limited to Tahiti, Viti Levu, Upolu, New Caledonia, New Zealand, and Australia, while Mr. Price, over a series of years, has made collections from Lord Howe Island, Norfolk Island, Rarotonga, Upolu, Savaii, several islands of Tonga, New Hebrides, Viti Levu, Lau Archipelago, New Caledonia, and several parts of Australia. These collections are housed in Field Museum. The Lau collections were received too late to be other than cross-referenced in this study. The non-Polynesian collections have provided exten- sive comparative material on which most of the higher classification decisions have been based. The rest of the material studied consists of type specimens, remnants of the Garrett, Mousson, and Ancey collec- tions, plus an occasional previously unstudied field collection and the subfossil to fossil species recorded by Ladd U958, 1968) and Aubert de la Rue and Soyer (1958). METHODS OF ANALYSIS Island taxa bring both advantages and dis- advantages to evolutionary studies. Particularly wher- ever the rate of immigration and species turnover is low and the opportunity for in situ speciation is thus enhanced, the possibilities of detecting convergences in structure between taxa on different islands or island groups are relatively good. Continental areas have been subject to the repetitive ebb and flow of faunal migrations, particularly in regions subject to the effects of the Pleistocene glaciation. Faunas of these areas usually are composed of multiple introductions, while the more isolated islands may have virtually monophyletic adaptive radiations. As is discussed elsewhere (Solem, 1973e), the Pacific Island land snails give evidence of low migration and high local speci- ation rates. Several situations were encountered where species from widely separated archipelagoes appeared virtually identical in many shell features. Deeper analysis of the structures and how they are formed enabled recogniz- ing that these similarities were convergent, although phenetic clustering techniques grouped them together. In several ways this study departs from the normal systematic criteria and methodology as cur- rently applied to mollusks. Far more statistical analysis of variation is utilized; an attempt is made to specify the philosophy used in recognizing species; and discussion of variational trends within generic units focuses on departures from an "ideal generalized morphotype" of that unit. Explanation of these procedures is a necessary prelude. VARIATION IN ADULT SHELLS Few people have attempted to undertake statisti- cal analysis of variation in shells of molluscan species that do not show determinate growth, such as usually is indicated by a reflected lip and/or denticles. It has been assumed that growth would continue throughout life and hence age factors would seriously bias or obscure any pattern to size or shape variation. In the zonitoids "...growth is usually persistent; that is, most species become sexually mature when comparatively small but continue to add whorls almost indefinitely" (H. B. Baker, 1938b, pp. 5-6). Thus Baker picked certain standard whorl counts and listed the size reached at these marker points to indicate size differences between species. Early in this study it became evident that postembryonic shell growth in the endodontids consists of two phases. For the first few whorls there is continuous growth, but then an alteration in pattern occurs that affects several features. In most taxa these changes are highly correlated, begin abruptly, and could be easily identified even by new assistants after an hour or so of instruction. These changes involve the pattern of rib deposition, umbilical decoiling, lip callus formation, and body whorl des- cension. The ribs become more crowded and/or irregular in spacing, often are reduced to a point at which they cannot be counted, and the microsculpture essentially disappears. The umbilicus becomes notice- ably wider for a period of growth, then the umbilical lip suddenly may reflect slightly in toward the center of the umbilicus. At the same time, the inner lip margin, particularly on the columellar wall, becomes greatly thickened, frequently with a thick callus, whereas in juvenile shells it is thin and sharp edged. The edge of the parietal callus may thicken and the apertural barriers become slightly wider at their anterior end. Finally, starting with the change in rib deposition, there is a tendency for the body whorl to descend more rapidly. These changes are obvious in most species. A notable exception involves those species whose growth pattern results in a small umbilicus, reduced sculpture, and thick apertural callus at all stages. In these species the juvenile growth is virtually the same as adult growth. Not every species shows all of the changes cited above and some alterations may occur before others. Never- theless, in only 1-2 per cent of the specimens is a judgment decision needed concerning whether a speci- men is showing juvenile or adult growth. The adult growth may occupy as much as one-quarter of a whorl, very rarely three-eighths of a whorl. There are considerable differences as to when it starts. Some individuals may start "adult" growth more than three- quarters of a whorl earlier than others. Such a change in growth pattern generally is correlated in animals with the onset of reproductive activity. Until that point, energy is channeled into increasing the size of the individual. Afterwards it is channeled into producing the next generation. Because the Bishop Museum mollusk collection is stored with the animals removed from the shell, it was not possible to study the timing link between reproductive matur- ity and these growth changes by dissecting individuals whose shell characters had been recorded. The only shell- associated endodontids available to me in quantity were Thaumatodon hystricelloides from 11 12 SOLEM: ENDODONTOID LAND SNAILS WHORL COUNT IN ADULTS 85- 80- 75- 70- 65- 60- 55- C/5 50- CO 40- m =,30- 25- 20- 15- 10- 5- 7 ' X WHORL COUNT FIG. 1. Frequency distribution of whorl counts in adult Libera fratercula from quantitative samples. Western Samoa, a species that shows very few adult changes, and Libera fratercula from Rarotonga, in which the formation of a brood chamber in the shell after reaching reproductive maturity is yet another modification. Shells showing a "juvenile" pattern of growth can be segregated from those that have a presumed "adult" growth pattern. The significant question in terms of analysis is what distribution of meristic features is shown by the "adult" specimens. Data from a quantitative sampling of live Libera fratercula from Rarotonga have been summarized by Solem (1969a). In the present paper data are given on grouped variation in all live adults. Of some 623 living snails, 219 juveniles and 391 adults could be measured. The others showed repaired breaks in the shell or were otherwise unsuitable for analysis. Figures 1-4 graph the variation in whorl count, height, H/D ratio, and shell diameter, respectively. Except for the diameter, these agree with Quetelet's principle in distribution of the size classes. The diameter is affected by the fact that after the start of reproductive activity in this species, the snail grows and gradually narrows the umbilical opening to form a brood chamber. During this growth the shell diameter often remains stable. Hence the skewed nature of the diameter curve. The whorl count is symmetrical, but the height curve is skewed positively, reflecting the continued growth during adulthood to narrow the umbilical opening. Smaller samples of other species show the same pattern, that of a normal distribution with at most a slight positive skewing, particularly in regard to rib counts (p. 42). I thus assume that measurements of adult shells will show either a "normal" distribution or slight positive skewing and hence can be compared through standard statistical analysis. The absence of determinate growth does not seem to have affected the basic normal distribution of adult variation in measur- able and continuously variable shell characters, thus interpretation of variation is based heavily on statisti- cal data concerning adult variation. Nature of quantitative data presented Basic quantitative data are presented in two formats. First, there is a summary of variation in size, shape, and apertural dentition for each species of that cluster. These tables are headed "Range of variation in ." The second format records adult size and shape variation within either specific local populations (where such refinement of collecting data is available) or particular museum sets (either single sets or material originating from the same collection) in the case of specimens dating from the late 1800's. These tables are captioned "Local variation in ." Unless stated otherwise in the discussion of a particular HEIGHT IN ADULTS 95- 90- 85- 80- 75- 70- 65- 60- 55- 50- 45- 40- 35- 30- 25- 20- 15- 10- 5- Fio. 2. Frequency distribution of shell height in adult Libera fratercula from quantitative samples. METHODS OF ANALYSIS 13 species or set, all measurements were made only on adult and gerontic individuals. In the tables summarizing range of variation, certain items require explanation. The number of specimens examined does not refer to the number of measured adults, but rather to the total number of specimens, both adult and juvenile, available for this study. For the ribs, height, diameter, Height /Diameter ratio (hereafter H/D ratio), and Diameter/Umbilicus ratio (hereafter D/U ratio), the first figure is the simple arithmetic mean of all measured adults. This is followed in parenthesis by the range. For the whorl counts, the cited range omits + and signs. Where species have been collected again in recent years, these measurements summarize data from many snail generations apart. Under these circumstances, calcu- lation of variance would serve no useful purpose. Definitions of the apertural barriers are given elsewhere (p. 53). In the summary tables, Pr means parietal, C columellar, and P palatal. In each case, the numbers before the + sign refer to major barriers, the numbers following to minor traces. Where size reduc- tion is involved, recognition of this difference involves considerable subjectivity. Where such arbitrary divi- sions were made, they are discussed in the text. Variation in barrier numbers is indicated in two ways. Two numbers connected by a dash (i.e., 3-4) indicate DIAMETER IN ADULTS H/D RATIO IN ADULTS 90- 85- 80- 75- 70- 65- 60- 55- 50- 45- 40- 35- 30- 25- 20- 15- 10- 5- 65- 60- 55- 50- 45- 40- 35- 30- 25- 20- 15- 10- 5- T T 500 600 800 Fid. 3. Frequency distribution of Height/Diameter ratio in adult Libera fratercula from quantitative samples. FIG. 4. Frequency distribution of shell diameter in adult Libera fratercula from quantitative samples. that at least one-third of the individuals have the less frequent state. If one of the numbers is underlined (i.e., 3-4), this means that the underlined number is found in less than one-third of the individuals examined. Usually it means a rare to infrequent character state. Occasionally, a number such as .2_-3-4_ + 4_-6 may be seen. This is translated as usually 3 major barriers, rarely 2 or 4; normally with six accessory traces, rarely only four. In the tables documenting variation within local populations or single museum sets, the number of specimens refers to those actually measured. For the whorl counts, only the mean and range is given, since the measuring error is relatively large (p. 15). For the other parameters, mean, standard error of the mean, and range are given in that sequence, for example 4.32 0.036 (3.97-4.46). Sample bias The material in the Bishop Museum collected by Cooke and Kondo, plus the specimens in Field Museum of Natural History collected by Solem and Price, provide samples undistorted by subsequent museum activities. In the case of specimens from the Garrett and Mousson collections, we are dealing with small remnants of initially large sets. Specimens were traded in driblets of two or three to other collectors in exchange for species new to the trader. Those samples remaining in the Mousson and Garrett material are size biased. I have shown elsewhere (Solem, 1966b, p. 16) that such trading often will result in elimination of smaller shells from the sample, thus increasing the mean size significantly. This was especially noticeable in several species of Libera during this study, but even 14 SOLEM: ENDODONTOID LAND SNAILS in the case of the 3-5 mm. endodontids, trading seems to have resulted in size bias. Hence only very cautious conclusions have been drawn from comparison of samples collected during the 1800's with material collected during the 1930's. A seldom appreciated collecting phenomenon involves unconscious field bias towards selecting larger or "adult" specimens and neglecting the smaller, possibly juvenile individuals. This has been docu- mented by H. B. Baker (1962b, p. 22) for North American Allogona and Puerto Rican pomatiasids. I have experienced this personally in relation to Pan- amanian Mexcyclotus and Samoan Ostodes. It is seen in this study with sets of Libera bursatella orofen- ensis. This difficulty is greater in those forms that do not have a cessation of growth that is marked by reflected lip formation at "adulthood." A strong bias towards gerontic and adult shells with a distinct neglect of subadult and juvenile specimens is quite possible, particularly when secondary interest is at- tached to the group, as in collection of snails by botanists or herpetologists. Although they were not collected on a randomized or quantitative basis, material accumulated during the several Bishop Museum expeditions seems to have avoided such a bias. Most sets contain a broad mixture of age groups. I could detect very little evidence of selection for adults in the collecting process, and then only in species taken by non-malacologists. The collections were not quantitative samples, however, and only a few sets were large enough to infer data on population structure. The reason for the probable lack of sample bias stems from the purpose of the trips. The collectors were primarily concerned with obtaining members of another family, the Tornatellinidae (Kondo and Clench, 1952, p. 17). Even in this group, they were attempting to gather all specimens seen, rather than to take a "good sample" of the popu- lations. Since 1 ) the small endodontids previously were essentially unknown; 2) they are very difficult to distinguish without use of a microscope; and 3) no particular emphasis was placed on the family by the collectors, I am convinced that there is minimal bias in the material. Measurement reliability All measurements of the shells were made by me under a Leitz stereoscopic dissecting microscope, using an 8 X ocular, 1 X , 2 X or 4 X objectives and an ocular micrometer. The micrometer was calibrated with a stage micrometer at several different times during this study. No significant differences in the calibration readings were noted. Measurements were taken to the nearest 0.5 micrometer unit. For the different objec- tives, one micrometer unit equals: 0.131 mm. with the Ix objective; 0.0658 mm. with the 2x objective; and 0.0329 mm. with the 4x objective. Most specimens were measured using the 2 X objective. Sets with most or all specimens under 2.75 mm. in diameter were FIG. 5. Method of measuring specimens. A-B, shell diameter; C- D, shell height; E-F, spire protrusion; F-G, body whorl width; H-I, umbilical width. measured with the 4x objective and sets with most specimens over 5.00 mm. in diameter were measured with the 1 X objective. Accuracy of measurement was tested by remeasur- ing individuals in several small sets five times over a four-year period. No reference was made to previous measurements before or during each trial. It was possible to compare specimen with specimen rather than having to depend on means, since there was a METHODS OF ANALYSIS 15 wide size range within these sets. Reproducibility of the measurements varied with the parameter. Ba- sically standard measurements were used and taken as indicated in Figure 5. Greatest accuracy was obtained in measuring the diameter, least in making the whorl count. The diameter rarely (less than 5 per cent) varied more than 0.5 units in a measurement of 40-50 units or 1 unit in 80-90, an accuracy of 1.0-1.25 per cent. The more subjective height measurement showed a greater range of from 1 unit in 35-50 to 1.5 units in 65-70 an accuracy of 2-2.9 per cent with the heights measured in fewer units being less accurately recorded than those in larger numbers. Determination of umbilical width varied inversely with the size of the umbilicus the wider the umbilicus, the more accurate the measurement. For very narrow umbilici, those whose D/U ratio was more than 7, but less than 10, error was in the range of 8-12 per cent. For those with tiny umbilici, D/U ratio 10-30, errors of 25-50 per cent were present, since often only one or two units were measured with an error of 0.5 units. The wider umbilici, D/U ratios 2-4, could be measured to within a 2-5 per cent error. Rib counts (fig. 32) were accurate usually to within one rib in 50, two in 100 and up to six for over 200 ribs generally a 2 per cent error except where the ribs were very fine and numerous, which often resulted in a 3 per cent error. Most of these errors resulted from judgment decisions concerning rib status at repaired breaks or on gerontic growth sections. Whorl counts normally showed a Vs error, with estimated readings made to the nearest one-eighth. For this reason, no statistical treatment of whorl counts was attempted. Usually species differ- ences in this category were either quite pronounced or not noticeable. The margin of possible error in the basic measure- ments naturally affects the accuracy of calculated ratios. Particularly in regard to the higher D/U ratios, the actual numbers become virtually meaningless. These figures only serve to indicate the small nature of the umbilicus and are worthless for statistical comparisons. Use of the lower D/U ratios is possible with a higher degree of confidence. Calculation of D/U ratios, using the margins of error cited above and arbitrarily selected figures, produced a range of possible error from 6.5-6.8 per cent in the 2-4 D/U ratios and from 7.6-14.3 per cent in the 7-10 D/U ratios. Calculation of sample H/D ratios produced a possible range of error between 3.3 per cent and 3.8 per cent. It must be stressed, however, that these margins of error represent the calculated extremes. It was not thought worthwhile to determine the mean margin of error in measuring individual specimens, but for three samples that had been measured several times, I computed the means for H, D, H/D, and D/U. With 7 to 11 specimens involved, the sample means varied by: H - 0.4-0.7%; D -0.1-0.3%; H/D - 0.5-1.0%; and D/U - 1-2% for ratios less than 8. In making comparisons between samples, the above figures have been given considerable weight in deciding whether calculated differences are significant, or represent random variation. CRITERIA FOR SPECIES RECOGNITION Recognition of species on the same island presents a different problem from dealing with allopatric populations on different islands. Closely related sym- patric species must have "species recognition" features to maintain genetic integrity. These differ with the group of organisms. Where congeneric species of endodontid snails are sympatric to the extent of living under the same log, character displacement in termi- nal genital structures has occurred (tables LII-LIV). One repetitive type of such character displacement is for the normal pattern of two penial pilasters that are equal in size to become altered. One of the species will have the conservative pattern, while in the other, one pilaster will be greatly reduced. Examples of this are seen in Libera micrasoma (fig. 171h, two simple pilasters) and L. cookeana (fig. 172a, two unequal pilasters) from Station 865 on Tahiti and in most of the Rapan Opanara where there are extensive and complex sympatric-allopatric species groupings. An interesting variation is seen in the Marquesan Taipi- don, where T. fragila (fig. 138f) and T. varidentata (fig. 138h) on Hivaoa and T. centadentata (fig. 139f) and T. semimarsupialis (fig. 139b) on Nukuhiva are sympatric or the handful of known specimens have been collected on the same ridge in different years within 90 ft. recorded elevation. In each case the first species has the pilasters broken up into a series of longitudinally arranged tubercles, while the second species retains the "typical" pattern for that genus. In the Nukuhiva Planudonta intermedia (fig. 147b) the 4.6 mm. long penis has two pilasters that are almost circular in cross-section, while the probably sympatric P. concava (fig. 147d) has one pilaster of the 6 mm. long penis altered into a high folded ridge. Where more than one genus from a monophyletic assemblage and several species are sympatric, as on Rapa Island, there are changes in size of penis, number of pilasters, shape of pilasters, and the length of the penial pilasters. Data on these alterations are given in the generic discussions and in Tables LII-LIV. They are not repeated here. What is basic is the concept that endodontid land snails use penial surface features in species recognition. When sympatric populations show dimorphic penial surfaces, I conclude that they are distinct species. In all such cases discovered so far, the differences in penial structure correlate with conchological differences. The differences may be large and obvious, as between Taipidon centadentata (fig. 144) and T. semimarsupialis (fig. 143e-f) or Plan- udonta concava (fig. 149a-b) and P. intermedia (fig. 149c-d); or the differences in shell features can be relatively small, as in Taipidon fragila and T. 16 SOLEM: ENDODONTOID LAND SNAILS uaridentata (fig. 145). Although the size of the morphologic gap in shell features varies, these differ- ences correlate with the penial gap and make it possible to sort the station material into two species. The penial differences indicate recognition of genetic incompatibility by the snails. The conchological differ- ences are expressions of underlying genetic shifts that led up to the point of genetic incompatibility. Anatomical study of sympatric material thus made it possible to recognize which snails consider each other different species. Analysis of shell variation provided greater to lesser correlated shell changes by which the species could be identified. When populations are allopatric because of a distinct water gap between islands or from living at opposite ends of a large island, there will have been no selective pressure for "species recognition" characters to evolve. If these isolated taxa have not been sympatric with congeners or closely related genera, then selection probably would be toward a stabilized penial surface. Any deviation from a "normal" pattern would be at a selective disadvantage, and would tend to have less reproductive success. In all probability the extremely conservative pattern of penial structure in the family relates to just this situation. At the same time, the populations will have been subject to varying selective pressures in their different habitats. Genetic change in other systems having adaptive value to local conditions will be selected. In keeping with the general pleiotropic nature of muta- tions, such changes can be expected to be expressed in shell feature shifts, even if the particular selective advantage of the change does not involve the shell directly. If these allotropic populations with different genetic systems but virtually identical penial surfaces became sympatric, either the genotypes would mix (if genetic compatibility existed) or rapid selection for species recognition characters would occur (if cross- population matings produced reproductively sterile or no offspring). It is obviously impossible to test the breeding potential of the many allopatric populations with similar penial types. In judging their reproductive potential it is necessary to make predictions on the basis of observations concerning sympatric popu- lations. The extent of conchological variation within populations can be measured and the minimum degree of conchological difference between sympatric species determined. In the vast majority of cases the degree of difference will exceed such minimums. There is no reason to assume that the currently observed min- imum represents the actual minimum necessary for genetic incompatibility to exist. Nevertheless, this is the best available starting point for evaluating the significance of observed differences between popu- lations. It is based on the premise that genetic divergence will be reflected by phenotypic alterations in the shell. When the genetic systems have diverged to the point of reproductively significant in- compatibility, then, if sympatry occurs, selection for species recognition characters on the functioning surfaces of the penis will take place. After the minimum conchological difference for a sympatric pair of species in the family is determined, the relationship of allopatric populations is judged against this standard. If the degree of difference is less than the standard, either clinal variation or subspecific differentiation is proposed. If the degree of difference is equal to or exceeds the standard, then the populations are assigned to different species regardless of the similarity in "species recognition characters." A precise definition of "minimum conchological difference" is difficult because shells can vary in many ways and different taxa show different sets of variations. Probably Taipidon fragila and T. vari- dentata show the least substantial changes of any absolutely or near sympatric species pair where it was possible to dissect both species. The former (fig. 145) has a flatter spire, narrower umbilicus, finer ribbing with fewer microradials between, and a greatly reduced number of apertural barriers. These are uncorrelated characters in this situation, since flattening the spire normally will widen the umbilicus, while finer and more widely spaced ribbing will have more microradials between each pair of major ribs. Taken singly, each of these characters can be shown to vary within a species unit as widely as they do between the two Taipidon. Ruatara oparica from Rapa has the nominate race (fig. 113c-d) with prominent apertural barriers, a subspecies R. o. reductidenta (fig. 114e) with greatly reduced barriers, and a third geographic race, R. o. normalis (fig. 114a, c, d), with an intermediate condition. Races of Opanara areaensis from Rapa (fig. 104) differ in spire protrusion or ribbing and umbilical size, but retain anatomical features in common. It is the combination of uncorrelated shell and penial differences between the two Taipidon that is significant. I am reluctant to establish a minimum number of uncorrelated differ- ences to be used as a fixed criterion for species separation. Instead I have given conscious weight to the apparent "key character usefulness" of a character within the context of that section of the family in making particular decisions. Again using Rapan taxa as examples, it is instructive to examine two other situations where judgment had to be made concerning the status of allopatric populations. Two pairs of taxa in Opanara that consist of essentially single relict populations (fig. 100) show virtually equivalent shell differences. Table VI summarizes morphometric data. The differences between the races of Opanara megomphala (Fig. 106a- d) stem primarily from a single change in spire elevation that affected the H/D ratio and D/U ratio, but had little effect on the diameter and ribbing. Because there was no noticeable difference in the METHODS OF ANALYSIS 17 TABLE VI. - SIZE AND SHAPE VARIATION IN RELICT, ALLOPATRIC RAPAN OPANARA 0. m. megomphala 0. m. tepiahuensis 0. caliculata Diameter H/D ratio Whorls D/U ratio Ribs on body whorl 3.21(2.83-3.52) 0.511(0.1*86-0.561*: 5 l/U-6 1/8 2.22(2.15-2.35) 73.7(71-76) 0_. altiapica 3.36(2.98-3.77) 3.27(3.09-3.39) 2.82(2.63-3.03) 0.1*3 1 +(0. 353-0. 1*81) 0.617(0.595-0.660) 0. 719(0. 66*-0. 761: 5 1/2-6 1/8 1* 5/8-5 1/1* 1+ 7/8-5 5/8 1.95(1.72-2.08) l+.7l*(l*.09-5.22) 1*. 69(3. 91-5. 53) 6l*.6(5l*-8l) 120.3(117-125: 80.0(61*-9l) genitalia and the statistically significant differences were the result of a single alteration, the populations are called subspecies. Opanara altiapica and O. caliculata differ in rib spacing (fig. 105a, c), coiling of the last whorls, size, and penis structure. They are called different species because there were several uncorrelated shell characters in addition to the penis "key," although the degree of statistical difference in respect to measured variables is no greater than between the races of O. megomphala. While each marginal case has to be decided on its own merits, where three or more substantial, non- correlated conchological differences exist between allopatric populations, and these changes are not known to differ significantly within populations in that section of the family or genus, then allopatric populations have been ranked as distinct species. If fewer than three such changes are present and there is no indication of "species recognition" differences in the terminal genitalia, then normally less than specific level differentiation is proposed. Subspecies have been recognized in two situations: where there are one or two uncorrelated characters that shift with at most slight overlap between populations on different islands or where there are dramatic "step dine" patterns of change involving adjacent areas of the same island. The few subspecies delineated in this study are true "incipient species." Further work might demonstrate that they have reached the level of being distinct. Except for the populations of Orangia cookei, Opanara areaensis, and Ruatara oparica on Rapa, natural test of sympatric or near sympatric populations occurrence does not exist. The decision on species level versus subspecies level designation has been, of necessity, subjective. For the Rapan taxa, the populations are nearly sympatric and the absence of "character recognition" features has led to the subspecies-level designation. As indicated in Table VI, the degree of measurable difference may be just as large between some of the subspecies as it is between certain sets of species. For this reason, species and subspecies have been given equal status in analyzing patterns of species level variation. NATURE OF COMPARATIVE REMARKS If comparisons are being made between the species living on a single island or at a single station, where either there are character displacement phenomena in respect to closely related species, or the species come from a variety of phyletic lines, then both quantitative and qualitative distinctions usually are obvious. Most users of this monograph will be attempting to identify species from a particular island, and hence for ease both in construction and user convenience, most keys are based on geographic units. A meaningful and usable comprehensive key to generic units would require extensive use of anatomical features observable only by dissection, since conchological variation is sufficiently large to require multiple entries and a highly artificial arrangement. Within genera, keys are based on the known species and represent artificial identification aids, not an attempt to depict phyloge- ny. Again, to be of maximum usefulness to the reader, they are based on conchological data. When comparing generic-level taxa, it has been possible, in most situations, to use easily observable and sharply distinct features. To some extent such remarks are based on departures from idealized basic morphotype for the family. The various genera are viewed as progressive departures from a basic pattern represented among extant taxa by Minidonta and Cookeconcha. Such patterns of change have been somewhat repetitive, as recognized by the Mau- todontha, Nesodiscus, and Libera levels of special- ization. Hence generic comparisons are a combination of direct character differences and patterns of depar- ture from the more primitive conditions seen in less specialized, potentially ancestral genera. Within genera, the non-sympatric species may differ by combinations of characters that are hard to define precisely and that are not reflected by averages of the relatively crude shell measurements. Hence it 18 SOLEM: ENDODONTOID LAND SNAILS was decided, before writing the definitive descriptions, where a homeostatic genetic system has been bent in diagnoses, and comparative remarks, to determine the several directions under diverse environmental pres- average pattern within that genus, or in the case of sures, is implicit in most systematic literature, but genera with one to three species, to use the presumed frequently has not been articulated or rigorously ancestral group average as the standard of comparison. applied in analysis. This procedure tends to magnify As species depart from the average pattern, these the size of differences between species, but is a differences are commented on in the diagnoses and necessary facet of communication, remarks. This view of a genus as a monophyletic unit, PATTERNS OF MORPHOLOGICAL VARIATION Initial investigations were carried out using gross methods to study shell and details of anatomic features observable at 100 X magnification or less with a dissecting microscope. Slide mounts of unstained radulae and jaws were examined using a Leitz Ortholux compound microscope under bright-field, dark-field, and phase-contrast illuminations. At a later period it was possible to make some scanning electron microscope observations of shell structure and sculp- ture (pp. 30-41), apertural barriers (Solem, 1973b), and radular denticles (Solem, 1972a, 1973a). Histological and cytological investigations of anatomical structures were not attempted. The Bishop Museum anatomical material had been preserved according to the method of Cooke (Cooke and Kondo, 1960, p. 4). This involved drowning the snails for 12 hr, followed by storage in 50 per cent alcohol until they could be sorted in the laboratory perhaps weeks or months later. After sorting the material into species and age classes, the soft parts were extracted by a water jet or tweezers (see Kondo and Clench, 1952, pp. 27-28 for a description of the process). The soft parts often broke and only terminal portions of the genitalia were available for study in many species. After a short immersion in 95 per cent alcohol, permanent pre- servation of the soft parts was in 75 per cent ethyl alcohol. While this procedure often yields material that is quite satisfactory for dissection and gross study, specimens preserved in this way are not suitable for histological investigations. Sometimes there was dis- tinct tissue degeneration in the post-pallial systems, so that apical genitalia could not be studied. The following discussion surveys the extent and frequency of occurrence for patterns of variation found in various shell and anatomical features as observed under the limitations outlined above. Where necessary to understand the significance of certain features, contrasts are made with the same structures found in the Charopidae, but a detailed family-level comparison of shell structure is withheld until Part II. Some data on family level differences have been presented previously (Solem, 1973b). Statements concerning the direction of character change are presented below without direct justification. The rationale for deter- mining primitive and derived in relation to the Endodontidae is covered in the section on phylogeny and classification (pp. 102-116). Classification has been based on anatomical rather than conchological criteria, but the shell is the most usable guide to species identification. In view of the virtual extinction of the family, the shell is the only system that will be available for future study. Hence patterns of conchological variation are reviewed first, and phyloge- netic trends are discussed later. SIZE AND SHAPE VARIATIONS The following statistical calculations and most charts omit data concerning the Lau Archipelago species, Priceconcha tuvuthaensis and Thaumatodon spirrhymatum, described by Solem (1973d). Their inclusion would not have materially altered the results, but would have required repeating a great deal of work and revising figures for little purpose. For each species and formally delineated sub- species, means and ranges of the basic measured parameters were calculated and then, together with other variational data, coded and key punched onto IBM cards for analysis. While these data indicate a few general patterns of change and permit some statements concerning the effect that various struc- tural alterations have on gross measurements, these averaged data must be used with caution. The extent of variation within members of a population in regard to shell shape and form is far greater than would be suggested by the computer-simulated shells of Raup (1962) or the studies on Poecilozonites and Cerion by Stephen J. Gould (1969, 1971). Particularly in regard to variations in ratios reflecting umbilical size and shell form, individual specimens of endodontoid species show little close linkage of variables. The endodontids have a messy, highly varied, and frequently individ- ualistic "generating curve" that frequently changes drastically and perversely during ontogeny. A prime example of this is the formation of a "brood chamber" or "brood pouch" from the shell umbilicus by secondary inward growth of the umbilical lip over a short segment of a whorl to more than a full whorl of shell growth (pp. 27-30). In Anceyodonta (figs. 81c, 82f) there are the rapidly changing umbilical contours. Major spire protrusion alterations can be seen within populations of one species, Nesodiscus taneae (fig. 152), and between subspecies of Opanara areanensis (fig. 104a, c, e). Zonitoids, in contrast, are far more regular in growth form. While minimum, mean, and maximum adult dimensions were recorded for each taxon, graphing of the minimums, selection of which involved judgment 19 20 SOLEM: ENDODONTOID LAND SNAILS mean freight maximum heigtil FIG. 6. Frequency distributions of mean and maximum shell height in the Endodontidae. greater whorl count. Most endodontids lay their eggs in the shell umbilicus, but on Rapa and Mangareva Islands there was a tendency for closure of the umbilicus. Figure 10 charts, on a log scale, the mean Diameter/Umbilicus ratio for all taxa. Figure 11 separates the Mangarevan and Rapan taxa for comparison with the overall pattern. Except for the two subspecies of the Palau Island Aaadonta fuscozo- nata (Beddome) (fig. 206), barely perforate or closed umbilici are limited to species from Rapa and Mangareva. For those species in which the number of ribs on the body whorl can be counted, the frequency distribution of rib counts is relatively symmetrical (fig. 12), although the pattern of rib spacing is more strongly skewed (fig. 13). = 20 minimum H/D 'atio --- mean HID ratio maximum HID ralio 2 93 343 405 481 5 63 6.61 781 941 11D! 12 50 :75 325 375 425 4(5 S25 ' 62S 675 '25 775 825 875 925 H/D ratio FIG. 8. Frequency distributions of minimum, mean, and maximum H/D ratios in the Endodontidae. FIG. 7. Frequency distributions of mean and maximum shell diameter in the Endodontidae. Diameter intervals are arithmetically equal segments from a logarithmic scale. decisions, is not presented for basic parameters although it is for most ratios. The pattern of shell height (fig. 6) is positively skewed in distribution, resulting from the evolution of species with a secondary brood chamber. As the shell continues to grow and narrow the chamber opening, height increas- es at a far greater rate than do other parameters. The growth vector alters to increase the height, whereas the diameter of these species is but little affected or even remains stationary. The diameter plot (fig. 7) is on a log basis and is symmetrical. As would be expected, the Height/ Diameter ratio is slightly skewed, (fig. 8) reflecting the height asymmetry. Whorl count (fig. 9) correlates well with the height pattern (fig. 6), again reflecting the number of species with a "brood chamber," since the increased height results from a man i mum . . \ '-I 375 425 475 525 U5 625 675 725 IT) 825 875 925 whorls FIG. 9. Frequency distributions of mean and maximum whorl counts in the Endodontidae. PATTERNS OF MORPHOLOGICAL VARIATION 21 45 40 35 30 20 10 minimum D/U mean D/U maximum D/U . I I I 1.50 1.86 2.31 3.56 4.42 5.49 6.76 D/U ratio 8.41 10.51 13.01 16.41 20.31 30.00 b o a pe e rl y closed FIG. 10. Frequency distributions of minimum, mean and maximum D/U ratios in the Endodontidae. D/U ratio intervals are arithmetically equal segments of a logarithmic scale. Each of the above basic measurements results from the interaction of several variables. Factor analysis would permit assigning weight to individual elements, but this would require far more elaborate and accurate measurements prepared from shell cross- sections than seemed worthwhile for this systematic review, plus computer competence and access that were not available to me. As examples of the changes, discussion is presented here concerning the effects of two more obvious changes, body whorl contour and spire protrusion, in relation to the basic measurements. Then a brief review on the pattern of change in umbilical contours precedes consideration of brood chamber formation and its effects on shell size and shape. A short concluding section on whorl count correlated variation provides background data for subsequent phylogenetic analysis. Body whorl contour While no quantitative measurement of body whorl contour was possible, visual coding into five categories was relatively unambiguous. The actual coding of the states was done from specimens, the written diagnoses, and illustrations, with difficult judgment decisions necessary in only a few cases. The approximate contour states are indicated in Figure 14A-E. These are termed: A) laterally compressed; B) evenly rounded; C) flattened slightly above and below a 22 SOLEM: ENDODONTOID LAND SNAILS 45 f I I I 1.75 3.00 4.00 5.00 8.00 1725 D/U ratio brood pooch FIG. 11. Pattern of umbilical size in Rapa and Mangareva Island Endodontidae compared with the frequency distribution for the entire family. rounded periphery; D) flattened above and below an angled periphery; and E) distinctly keeled. A sixth category, F, contains all species in which a brood chamber has been developed. The latter change overrides the alterations in whorl contour. Treatment of these taxa separately is essential. In the four charts, the x axis digits refer to the number of taxa in that grouping, while the y axis figures give the measurement range for the particular parameter. There is no correlation between D/U ratio (lower right of fig. 14) and body whorl contour variation, and virtually no change in shell height (upper left) until a protruded keel is developed. The development of a peripheral keel reduces the options for attachment of succeeding whorls. To fasten the parietal-palatal margin above such a keel leaves the parietal wall with a strong projection into the very area of the apertural cross-section where the comparatively bulky heart and kidney lie and through which (at the apertural opening) the head with its bulbous buccal mass must be withdrawn. Such a procedure would be highly inefficient. I know of no snail in which the attachment lies well above a protruded keel. The peripheral keel itself, or even slightly below it, marks the attachment point for subsequent shell growth. The slight height increase shown by keeled species probably reflects a negative feature, in that once the keel is started, reduction in S, 20 minimum ribs mean ribs maximum ribs 70 90 110 ribs on body *tK>rl 130 150 170 190 200' FlG. 12. Frequency distributions of minimum, mean, and maximum rib counts on the body whorl in the Endodontidae. Species with reduced or irregular ribbing omitted. shell height becomes mechanically far more difficult to accomplish. Brood chamber species (state F) are included to emphasize their increased shell height. Shell diameter (upper right of fig. 14) increases steadily with the change in whorl contour, except for going from state C to D. Since the only alteration in this transition involves the upper and lower lateral margins of the body whorl, the periphery-to-periphery distance would remain unchanged. In state E the 401- 135 7 9 11 13 15 17 19' ribs per mm. FIG. 13. Pattern of rib spacing on the body whorl in terms of ribs/mm, of peripheral circumference. Species with partly reduced or irregular ribbing omitted. PATTERNS OF MORPHOLOGICAL VARIATION 23 median x height E E 38 43 28 29 diameter 38 43 21 14 28 29 H/D ratio D/U ratio 060 r 055 050 38 43 28 29 6r- 37 37 19 n 13 27 FlG. 14. Effects of changes in peripheral whorl contour on some basic measurements in the Endodontidae. See text for explanation of character states A-F, "n" refers to the number of species level taxa having each state. The D/U ratio graph omits species with closed or barely perforate umbilici. actual lateral protrusion of the periphery on each side would, and does, drastically increase the diameter. The proportionate increase in diameter from states E to F is less than the change in height (upper left). Again this reflects the mechanics of brood-chamber forma- tion. Correlated with the increase in shell diameter, which results from further lateral expansions of the body whorl, is a gradual decrease in the H/D ratio (lower left). The slight mean levelling in the H/D ratio of state E and rise in height for group E results from the Palau Island Aaadonta being mostly in this group. This genus has by far the most elevated spire and greatest number of species with keeled peripheries for any genus in the family. Data on whorl count correlations are presented in Table VII. The difference from state C to D is not significant ("t" = 1.3373 with 33 df), but the change from C to E is significant ("t" = 2.6954 with 47 df), It may be that the increase in shell diameter combined with keeling is a mechanism that serves to decrease shell-height increment as the whorl count is raised. A retardation of the increase in shell height would keep open narrower niches for the snail to retreat into. The keel protrusion would adjust cross-sectional whorl area to compensate for the lessened height increment. In summary, changes in the whorl contour directly affect the shell diameter and H/D ratio, but have no correlation with D/U ratio and only slight direct effects on shell height. Whorl-count increase correlates slightly with keel formation, but otherwise there are no major identifiable results from whorl contour changes. 24 SOLEM: ENDODONTOID LAND SNAILS TABLE VII. - WHORL CONTOUR AND WHORL COUNT CORRELATIONS Character state A B C D E F X and SEM whorl count 5. 5.350.11 5.350.12 5.66+0.21 5.8?0.lH 6.78+0.11 Number of taxa 38 21 1U 28 29 Spire protrusion If the whorls of a typical shell could be unwound and then coiled in different fashions, certain changes would be obvious. If the rate of whorl descension was great, resulting in a protruded spire, and the whorl count remained the same, the shell height would be larger, the diameter less, the H/D ratio would approach nearer to unity (or even above), the umbilicus would be narrower, and the D/U ratio larger in number. If the spire was flattened or depressed resulting in little or no whorl descension, the height would be less, the diameter greater, the H/D ratio a lesser proportion of unity, the umbilicus wider and the D/U ratio lower in number. Isolating spire protrusion for analysis is difficult, since the degree of body-whorl deflection during adult growth varies greatly between taxa. This parameter has great influence on shell height and some influence on shell diameter. To use the ratio of spire protrusion above the body whorl to total shell height would add the variable of body whorl descension. I have chosen to indicate the degree of body whorl protrusion through use of an index obtained by dividing the actual spire height above the body whorl (E-F in fig. 5) by the body whorl width (F-G in fig. 5) taken at a point directly below the spire. Adjustments in the latter measurement to allow for the changes resulting from keel protrusion would slightly refine the data, but were not feasible or thought necessary at this level of analysis. Measurements were made on type specimens only, with the exceptions noted below. The resulting data were clumped into five states, which are characterized as: A) spire flat or depressed; B) index 0.01-0.250; C) index 0.251-0.500; D) index 0.501-0.750; and E) index 0.751 to the observed maximum of 1.33 in Gambiodonta grandis. About 10 per cent of the species were sufficiently near the group dividing points that several specimens were measured and the means used to decide group placement. Individual variation within large populations was far less than found in the Charopidae, although the total range frequently overlapped two states. The vast majority of taxa could be assigned to a group without question. A rough indication of the divergent appear- ance is given by the small diagrams at the top of Figure 15. Both mean and median parameters are given in the diagrams. The major effect of spire elevation is on H/D (lower left) and D/U (lower right) ratios. As the spire is protruded, the H/D ratio increases and the umbilicus becomes narrower. Two minor shifts in the charts require explanation. The lack of H/D ratio change between C and D probably reflects the fact these ratios are the typical pattern for the family (fig. TABLE VIII. - SPIRE PROTRUSION AND WHORL COUNT CORRELATION Character state A B C D E Brood chamber absent Brood chamber present Whorl count X and SEM 5.21+0.19 5.2^+0.12 5.52o.09 6. 03*0. 12 6.o8o.i6 Number of taxa IT 59 21 5 7.63 6.98 6. hO 6.93 Number of taxa 1 h 10 PATTERNS OF MORPHOLOGICAL VARIATION 25 median x height diameter 3.90 r - 3.15 Z40 1.65 17 44 63 31 n 6.50 r- 5.50 4.50 350 19 17 44 63 n 31 19 H/D ratio 0.650 0.600 0.550 0.500 0450- 9.00 7.50- 600 450- 3.00 D/U ratio without brood pouch 17 44 63 n 31 19 17 41 63 n 19 FIG. 15. Effects of changes in spire protrusion on basic measurements in the Endodontidae. See text for explanation of character states A-E, "n" refers to the number of species level taxa having each state. 8). Either a more or less protruded spire represents a departure from the norm and hence may drastically alter the H/D ratio. In the D/U ratio chart, the dip in the group E median results from inclusion of the very widely umbilicated Nesodiscus in this category. Changes in shell height (upper left) follow the common sense evaluation of the data, as essentially do the changes in shell diameter (upper right). If the spire is pushed down and into the middle of the coiling plane, then in order to preserve the same total whorl volume, diameter would have to increase; thus the greater diameter for the group A taxa. Flat or depressed spires apparently are advanced characters in the Endodontidae, thus part of the diameter increase probably is phylogenetically determined. The greatly increased diameter in groups C, D, and E reflects two factors an increasing percentage of taxa with brood 26 SOLEM: ENDODONTOID LAND SNAILS chambers developed and an increase in whorl count. This is documented in Table VIII, which shows that the development of a brood chamber correlates with a moderately high spire, but there is no correlation among brood pouch species between spire protrusion and whorl count. Umbilical contour Umbilical openings are used by endodontids as an egg deposition site. The variations in umbilical contour and basic shape are far greater than in mollusks where this opening is a byproduct of growth and lacks other functional significance. In discussing the taxa, the umbilicus was described as showing one of 11 shapes and contours. These were subsequently coded as States: 1. V-shaped, regularly decoiling (40 taxa) 2. V-shaped, last whorl decoiling slightly more rapidly (3 taxa) 3. V-shaped, last whorl decoiling much more rapidly (2 taxa) 4. U-shaped, regularly decoiling (23 taxa) 5. U-shaped, last whorl decoiling more rapidly (22 taxa) 6. cup-shaped (18 taxa) 7. U-shaped, barely decoiling (17 taxa) 8. secondarily narrowed to form a "brood cham- ber" (29 taxa) 9. barely perforate (10 taxa) 10. closed by contraction (3 taxa) 11. closed by reflection of the columellar lip (7 taxa). Rough cross-sectional views and an indication of the multiple directions in character change are shown in Figure 16. The difference between V and U shaping is a factor of early ontogeny. In those with a V-shaped umbilicus, the initial umbilical width is narrower. During subsequent growth the point of attachment for the columella on the basal margin of the preceding whorl remains the same. The shape of the opening is a simple V. A good example of this is seen in Minidonta hendersoni (fig. 63d). In those with a U-shaped umbilicus, the initial umbilical width is proportionate- ly wider and the point of attachment for the columella on the preceding whorl is much nearer the basal- umbilical margin. The same point of attachment is maintained through subsequent growth. The basic shape thus becomes a U with only slight to moderate divergence of the sides as in Endodonta ekahanuiensis (fig. 166c). The difference in appearance is caused by both a different initial width of the umbilicus at the apex and changed attachment point for the succeeding whorls. There is no strict taxonomic linkage, since the two types of umbilici occur in equal numbers of Thaumatodon and Mautodontha, but Cookeconcha, Taipidon, and Australdonta have mostly V-shaped umbilici, while Opanara and Endodonta show mostly U-shaped variants. Table IX summarizes the size and shape correla- tions with umbilical contour shifts. The difference in whorl count between States 1 and 4 is significant ("f = 2.9974 with 61 df), but the other changes are not statistically meaningful. Taxa with U-shaped umbilici, States 4, 5, and 7, do show differences from each other. Those in which the umbilicus is regularly decoiling (State 4) have an obviously wider umbilicus and significantly lower H/D ratio ("t" = 2.0833 with 38 df) than those with a barely decoiling umbilical opening (State 7). Those in which the last whorl decoils more rapidly (State 5) average smaller in diameter than the others, but the range in size is so great that the difference is not significant ("t" = 0.7851 with 37 df). Unfortunately, there is virtually no data available concerning egg size in the Endodontidae, so that the significance of the essentially identical D/U ratio for States 5 and 7 is uncertain. It may be coincidental. Umbilici that are cup-shaped (State 6) occur primarily in taxa at the Nesodiscus level of specialization. These taxa have a statistically significantly higher whorl count than State 7 ("t" = 2.2095 with 32 df), an TABLE IX. - SIZE AND SHAPE CORRELATIONS WITH UMBILICAL CONTOUR CHANGES Umbilical shape State 1 State 1* State 7 State 5 State 6 Mean and SEM Height Diameter H/D ratio D/U ratio Whorls 3.80+0.15 5.080.09 3.8U+0.17 5.570.15 5.79+.O.U1 5.6l0.15 Uo 1.850.ll 3.780.2U O.U890.010 23 1.960.17 3.950.38 0.5070.013 17 2.0lj0.12 3.72+0.25 0. 562+0. 02k 22 1.78+0.19 3.32+0.39 0.550+0.017 18 2.080.18 i+.8U0.27 0.1+290.012 2.170.09 6.150.19 PATTERNS OF MORPHOLOGICAL VARIATION 27 Brood Pouch li r:''i;:-"l 3: fl lip reflection constriction constriction lip reflection FIG. 16. Patterns of phyletic change in umbilical shape and size. obviously wider umbilicus, lower H/D ratio, and are larger in size. Table X shows the relative lack of strong correlations between degree of spire protrusion and umbilical contour, except for taxa with either a brood chamber (State 8) or barely perforate umbilici. Spire protrusion is symmetrically distributed in frequency (see Totals column of table X). There are some skewed correlations in contour. V-shaped umbilici are restrict- ed to forms with relatively low spires. This corresponds to the doming effect mentioned by Gould (1969, p. 432). The shift in frequency shown by States 4 and 5 may be partly caused by the same phenomenon, as is the restriction of perforate umbilici (State 9) to those taxa with elevated spires. In contrast, closed umbilici, those barely decoiling, cup-shaped umbilici, and those with the last whorl decoiling more rapidly do not correlate with spire protrusion. Inspection of Figure 16 shows that from the generalized conditions, specialized states can be achieved in several different ways. In the systematic discussions under particular genera, the patterns of such change are reviewed individually. Such data is based on gross examination. A more elegant analysis would have been possible by using cross-sectional profiles, but the time needed for such preparations was not available. One aspect that greatly influences umbilical volume is the interior wall contour. Frequently the sides of the umbilicus are flattened. No accurate measurement of this was practical and, although mentioned throughout the systematic sec- tion, no general analysis is presented. The above sketchy data on umbilical shape and contour is necessary as an introduction to the major conchological change seen in the family, formation of a brood chamber. It also serves to suggest some of the complexities concerning the umbilicus in the Endodon- tidae, nearly all of which relate back to its functional use for egg deposition. Brood-chamber formation Shell growth that secondarily narrows a widely open umbilicus to form a brood chamber has occurred at least five different times in the Endodontidae. Three genera, Pseudolibera, Libera, and Gambia- 28 SOLEM: ENDODONTOID LAND SNAILS TABLE X. - UMBILICAL CONTOUR AND SPIRE PROTRUSION Spire Elevation Depressed or flat 0.01-0.250 0.251-0.500 0.501-0.750 over 0.751 Totals 17 kk 63 33 19 Umbilical States 1-3^56 16 20 13 6 donta, contain nothing but species with brood cham- bers. In two other genera, single species (Endodonta marsupialis and Taipidon semimarsupialis) show the secondary narrowing. Other species of the same genera have a U-shaped umbilicus that would require only the secondary narrowing to form a brood chamber. An additional three species show slight umbilical narrow- ing during the last whorl of growth and hence form prototype brood chambers. Kleokyphus callimus (fig. 95c) from Makatea, Tuamotu Islands, Kondoconcha othnius (fig. 162c) from Rapa, and Thaumatodon euaensis (fig. 194c) from Eua, Tonga all are relatively narrowly umbilicated species that have last whorl constriction of the openings. Thus a total of eight lineages in the family produced species in which the umbilical openings have become secondarily narrowed to form an egg-holding cavity. The three genera in which all species show striking umbilical alterations are very similar in size and shape. The method of chamber formation is different in the three genera and independent deriva- tion of the chamber is certain. Pseudolibera from Makatea (fig. 168) has the narrowing process occur over VA whorls. After a period of stabilized umbilical width, there is gradual inward growth of the entire columellar wall that accelerates for one-half whorl, then finally stabilizes in the same position relative to the shell axis for the last one-quarter whorl of growth. In Gambiodonta (fig. 185) from Mangareva, the columellar-basal margin first becomes angulated, then keeled. Following this there is an abrupt one-quarter to one- third whorl growth toward the middle of the umbilicus, followed by three-quarters to two-thirds whorl growth with the columellar-basal margin retain- ing the same position relative to the geometric center of the umbilicus. Completing the one whorl of growth produces a brood chamber with sharply narrowed opening. In Libera from the Society and Cook Islands, closure is the result of gradual inward columellar growth over about two whorls of shell increment. The only exceptions to this pattern are seen in such 7 13 3 3 3 2 7 1 2 8 5 1 8 10-11 i U 10 lit 2 5 5 1 depressed species as Libera gregaria and L. recedens (fig. 175) from Moorea and L. streptaxon (fig. 179) from Tahiti, where parietal wall detachment beginning at the last whorl of growth initiates and provides most of the impetus for umbilical narrowing. In these species this alteration seems to be a compensatory adaptation to preserve maximum volume inside the umbilical chamber. The depressed spires act to reduce the volume of the chamber, but the shorter closure period effectively adds to the volume. Taipidon semimarsupialis (fig. 143e-f) from Nuku- hiva, Marquesas has an inward curve of the entire columellar wall during the last 1 1 A whorls of growth producing the brood chamber. In Endodonta marsu- pialis (fig. 167b) from Oahu, Hawaii there is an inward extension of the columellar-basal margin over the last two whorls of growth. Both of these species show many anatomical differences from the dissected Li- bera. Despite being unable to dissect either Pseudoli- bera or Gambiodonta, the differences in brood- chamber formation combine with other conchological criteria to separate them generically. Considering just the 29 specific-level taxa in Pseudolibera, Libera, and Gambiodonta plus Endo- donta marsupialis and Taipidon semimarsupialis, no brood pouch brood pouch FIG. 17. Mean shell height in species with and without a brood chamber. PATTERNS OF MORPHOLOGICAL VARIATION 29 there are marked average differences from non-brood chamber species. Since the numbers of taxa involved are quite disparate, the data for Figures 17-20 have been converted into per cent of taxa for ease in direct comparison. Shell height (fig. 17) is obviously much greater in brood-chamber taxa, although a few of the normal taxa are as high. The nature of these exceptions involves gigantism (Nesophila tiara), a combination of large size and increased whorl count (Nesodiscus magnificus, N. fabrefactus, Kleokyphus hypsus), or high-spired, proto-brood chamber taxa (Endodonta kamehameha, E. fricki). Shell diameter (fig. 18) is less sharply demarcated in brood-chamber taxa, since the pattern of growth in forming the chamber involves height more than diameter. Selection for ability to retreat into narrow niches probably influenced the diameter and H/D ratio (fig. 19) to a marked extent. Finally, the increased whorl count (fig. 16 19 23 21 32 36 II 52 61 1! it 102 118 mean diameter FIG. 18. Mean shell diameter in species with and without a brood chamber. 20) of brood-chamber taxa probably reflects the fact that brood-chamber formation is an improvement upon a functioning system. Endodontids obviously survive very well by laying eggs inside a "normal" umbilicus. Formation of the brood chamber permitted laying a larger number of eggs, increasing the per cent retained until hatching (particularly if the eggs are close to the chamber opening in diameter), and possibly reducing predation upon the eggs. Thus by continuing growth after reproductive maturity, the reproductive success rate would be increased. Hence the larger number of whorls involved in brood chamber formation might be compensated for by a reduction in energy directed toward egg production. Brood chamber narrowing occurs after there is already a sufficiently large umbilicus to hold eggs. Thus the average whorl count would be higher in brood- chamber taxa. In at least some species the opening to the brood chamber becomes so constricted that the hatched young have difficulty exiting. In Libera streptaxon (fig. 180) specimens normally show eroded margins where the young apparently have gnawed their way to the external world. Libera fratercula from the Cook Islands carries this procedure to the ultimate possi- bility. During brood chamber narrowing the apical soft i s i 20 no brood pouch brood pouch 325 375 .425 475 .525 .575 mean H/D ratio 625 .675 725 .775 FIG. 19. Mean H/D ratio in species with and without a brood chamber. parts withdraw from the early whorls. Instead of leaving this area vacant or filled with mucus deposits, as is common in multi-whorled land snails, it is filled with calcium crystals, effectively turning the shell apex into a solid rock. The largest specimens will have soft parts in the last four whorls of growth with as many as 4% whorls "evacuated" and closed up with calcium. The young gnaw their way up through the shell apex in order to exit (Solem, 1969a, p. 11, fig. 3). Since this species lives in the shore-line coral rubble tossed up by severe storms, calcium is super-abundant. The extra calcium needed to fill in the apical whorls is readily available and retention of the young for a longer period keeps them in a humid niche for a longer time. No data is available concerning the exact size of the young at time of exit from the brood chamber. 50 40 30 Z 20 10 no brood pouch brood pouch 4.9 5.6 mean whorl count 6.4 7.8 FlG. 20. Mean whorl count in species with and without a brood chamber. 30 SOLEM: ENDODONTOID LAND SNAILS TABLE XI. - WHORL COUNT CORRELATED SIZE INCREMENT Whorl Count Range Median 3.6-1+.95 ^. 55 5.0-5.5 5.25 5.55-6.0 5.TO 6.05-6.95 6.1+0 7. 05-8. 05 7.30 I'lumber of taxa 27 58 35 1+0 17 There is considerable variation in size of the young within the brood chamber, so that growth does occur after hatching and before exiting. Effects of whorl increment Although in the Charopidae there is considerable and obvious difference in the rate at which the whorl cross-sections increase in size, species in the Endodon- tidae have a relatively uniform pattern of change. Except for obvious cases of gigantism in which the size of the nuclear whorls is grossly enlarged, such as Nesodiscus magnificus and Nesophila tiara, size increase in the Endodontidae correlates with contin- ued additive growth that results in a higher mean whorl count. Spire protrusion, body-whorl descension, keel development, and relative cross-sectional area influence and modify the results, but basically simple whorl count increase results in most of the observed size increment. Data on this are summarized in Table XI. The whorl-count intervals are not equal in distance, since the clustering around the median whorl count of 5Vz + was so great that a division into two parts seemed advisable. The relatively small size increment between the 5.25 and 5.70 medians is an artifact of shortened interval rather than a change in pattern. Both diameter and shell height increase regularly and dramatically as whorl count is increased. There is only minimal change in proportions correlated with whorl count, since the alteration in the H/D ratio is negligible, As has been demonstrated above (pp. 21-27), there are other factors that have much greater effects on shell proportions. SHELL SCULPTURE Since the 1850's almost any small helicoidal land snail with numerous radial ridges and/or a reddish- brown or flammulated color pattern was placed in the "endodontoid" complex. This was in contrast to the smooth, shiny-shelled "zonitoid" taxa. Dissections have split off various paryphantid, pyramidulid (pleurodis- cid), strobilopsid, polygyrid, and camaenid taxa, but a core of sculptured litter dwelling taxa remain for consideration. X and (% of change from prior group) Height Diameter H/D ratio l.U6(-) 2.80(-) 0.1+97(-) 2.11(1*14.5%) 3.67(31.1%) 0.523(U. 7%) 2.3Ml0.9%) 3. 85(1+. 9%) 0.51+1(3.!*%) 3.10(32.5%) l+.8o(2l+.7%) 0.561+(l+.3%) 1+. 1*1(1+2. 3%) 6.88(1+3.3%) 0.556(1.1+%) Recently I have been able to initiate scanning- electron-microscope studies not only of sculpture features on the shell surface, but also investigations of the mechanics used to bond together the periostracal and underlying calcareous layers. There appear to be major differences in bonding mechanisms, whereas similar external sculpture on the spire and body whorl has evolved in unrelated lineages. The data on bonding mechanisms are too fragmentary for presentation at this time. The Pacific Island endodontoid family taxa do show clear differences in their relative use of calcareous versus periostracal sculptural components, the formation of sculpture on the postnuclear whorls, and in the pattern of apical sculpture. Discussion of basic sculptural structure is restricted to these aspects pending completion of more detailed investigations on layer bonding and details of surface sculpture in Australian and New Zealand taxa. Data is presented below on the nature of micro- and macrosculptural elements; family-level differ- ences; additive sculptural features; patterns of vari- ation in major rib numbers, size and spacing; the nature of changes correlated with rib spacing and rib reduction; and a brief hypothesis concerning the functional significance of shell sculpture in the endodontoid genera. Types and growth patterns Normal shell sculpture in the Endodontidae consists of major radial ribs plus a complex micro- sculpture that is barely visible at 80-100 X mag- nification. The major ribs are visible either to the naked eye or at less than 30 X magnification. The microsculpture can be analyzed with the scanning electron microscope (SEM) at 300-3,000 X mag- nification. These two elements of sculpture are additive, with the microsculpture continuing onto and across the major swellings. This is quite clearly shown in, for example, Rhysoconcha atanuiensis (the lower half of fig. 21b). There are four to six microradial riblets in the "trough" between each pair of major ribs. After the initial abrupt "rise" of the major ribs, several crowded microradials are clustered on the upper and PATTERNS OF MORPHOLOGICAL VARIATION 31 anterior (descending) side of each major rib (left side in cited figure). The spacing between these micro- radials obviously is much narrower than between the riblets in the "trough" area. Inspection of other species shows similarity in results, but considerable variation. Gambiodonta agakauitaiana (fig. 22d-e) has riblet crowding on the posterior (ascending or right) side of the major ribs. Australdonta raivavaeana (fig. 23a-b), which has relatively low and inconspicuous major ribs, shows no consistent pattern of riblet crowding on one side or the other of the major ribs, although usually one side is more crowded. Nesodiscus taneae (Garrett) (fig. 31c-d), where major ribs are greatly reduced, shows no change in riblet spacing. Several Polynesian FIG. 21. Sculpture of Rhysoconcha atanuiennis and Thaumato- don decemplicata (Mousson): a-b, Rhysoconcha atanuiensis. Station 367, Atanui Bay, Rapa. BPBM 140161. a, apex and early postnuclear sculpture (ca. 300x); b, suture between apex and first postnuclear whorl, note absence of microspirals on postnuclear (ca. 1,000 X); c, Thaumatodon decemplicata. Paratype. Nukufetau, Ellice Islands. FMNH 116990 ex Mousson coll. Charopidae (Solem, unpublished data) consistently had riblet crowding on the ascending rib side. Such spacing variation can be investigated by growth studies. In the land prosobranch family Diplommatinidae, A. J. Berry (1962) has shown that Opisthostoma retrovertens Tomlin usually added one rib and an interspace during each 24 -hr. period. Such growth could occur during both day and night periods, but usually happened overnight. Repair of shell breaks and drying of the habitat either interrupted growth or produced finer and more crowded ribs. When the snails were kept under conditions of high humidity and in total darkness, rib production increased to more than one per day. Obviously, it is not possible to transfer this observational data to the totally unrelated Endodon- tidae. Available data does suggest the relative timing pattern involved in major rib formation, although not the length of the cycle. With the assumptions that the microradials are deposited at equal time intervals and that control of their deposition is partly independent of horizontal anterior growth, then the spacing pattern can be modelled quite simply. For example, in Rhysoconcha there would seem to be a sudden upward surge of both inter-riblet conchin template and calcium deposition that results in an underlying major rib protrusion accompanied by a widening of the area between adjacent microradials. Following this there is FIG. 22. Sculpture in Gambiodonta: a-b, Gambio- donta mirabilis. Station 277, Ganhutu, Mangareva, Gambier Islands. BPBM 138981; a, apex and early postnuclear whorl (ca. lOOx); 6, detail of late apical sculpture (ca. 300x); c-e. Gambiodonta agakaui- taiana. Station 195, Rikitea, Mangareva. BPBM 138903; c, apical sculpture, note crystallization patterns in wear areas (ca. 300 x ); d, postnuclear sculpture, note microreticulation (ca. 100 X ): e, detail of postnuclear rib endings at periphery (ca. 300 X ). 32 PATTERNS OF MORPHOLOGICAL VARIATION 33 FIG. 23. Postnuclear sculpture of Australdonta raivavaeana. Station 674, Mt. Turivao, Raivavae, Austral Islands. BPBM 147529: a, portion of last two whorls, note grooved pattern in upper left (ca. 100 X ); b, suture between body and penultimate whorl (ca. 300X ); c, detail of suture, note changed subsutural riblets (ca. l,000x ); d, suprasutural sculpture (ca. 3,000x). a marked slowing in horizontal growth that results in several microradials being formed close together in the conchin template layer. Gradual return to a normal rate of horizontal increment and sudden lowering of the accretion plane to the whorl surface level starts a new rib interstice. Growth would continue at an even rate to the next surge point, starting the cycle over again. Gambiodonta and the Charopidae would appear to use the opposite tactic, with a slowdown in horizontal growth and resultant microradial crowding preceding the major rib peak. Acceleration of the horizontal growth rate would increase the distance between the microradials on the outer rib face before returning to the normal interval pattern. Forms with reduced ribs could have irregularity in surges produc- ing the resultant variation in Australdonta. The lack of change in Nesodiscus probably is a result of low surface relief. Figure 24 diagrams the pattern of these activities. The basic hypothesis that microradial construction occurs at regular time intervals, while horizontal growth is in irregular surges requires experimental testing. This theory does indicate a way in which the different riblet spacing observed could be achieved. 34 SOLEM: ENDODONTOID LAND SNAILS JUU a I I %U^mA^^ mill 1 1 1 1 1 1 1 1 1 r 1 1 M 1 1 1 1 1 1 1 FIG. 24. Patterns of radial surface sculpture in selected Endodontidae: a, Rhysoconcha atanuiensis, based on Figure 21b; b, Gambiodonta agakauitaiana, based on Figure 22d-e, c, Australdonta raivavaeana, based on Figure 23a-b, d, Nesodiscus taneae (Garrett), based on Figure 31c- d. Upper diagram indicates cross-sectional view of shell differentiating conchin (outer) and intial calcium (inner) layers; lower diagram plots horizontal distance between radial riblets. All figures greatly enlarged, but not to same scale The outermost periostracal layer of conchin, a sclerotized protein layer, acts as template for the initial calcium layer. Whether this periostracum is present or splintered off in dried individuals, the apical microsculpture appears the same at 100 X mag- nification. Even at 3,000 X magnification (for example, Minidonta hendersoni, fig. 25d) bits of flaking peri- ostracum are seen to conform exactly to the under- lying calcium layer. This is not correct in regard to the postnuclear sculpture, where both riblets and ribs may have high, lamellate periostracal extensions. In a very few species, notably Cookeconcha hystrix (Pfeiffer) and C. decussatulus (Pease) (fig. 26a-c), the periostracum has elongated "hairs" or "setae" produced. In Rhyso- concha atanuiensis (fig. 21b) and Australdonta raiva- vaeana (fig. 23a) the partly lamellar extensions and underlying calcuim layers of particular ribs are obvious. This means that while the major sculpture visible to the naked eye is composed mainly of calcium swellings, the microsculpture consists of a primary FIG. 25. Sculpture of Minidonta hendersoni. Station 254, Henderson Island. BPBM 149858: a, apical and early postnuclear whorls (ca. 300X ); b, postnuclear sculpture, note fine microspirals; c, apical sculpture, note wavy microspirals (ca. l.OOOx); d-e, details of apical sculpture (ca. 3,000 X ). 35 36 SOLEM: ENDODONTOID LAND SNAILS conchin layer underlaid, except for the periostracal fringes, by a duplicate layer of calcium. The fringes usually are extremely narrow. Probably the height of the underlying calcium riblet is a practical compromise between the need for a second line of surface irregularity in case of conchin layer loss and the desirability of keeping the vertical extensions narrow for as great a distance as possible, to retain a narrow edge when partial breakage occurs. Family-level differences Apical sculpture in the Endodontidae consists primitively of fine radial riblets and large single radial ridges combined with a very fine microspiral sculpture of "squiggly" cords. The latter look as if someone with a hangover had attempted, unsuccessfully, to squeeze a straight line of toothpaste from a tube. These microspirals are crowded, very narrow and seemingly independent of the radial sculpture. Typical in appearance is the apical sculpture of Mautodontha aoraiensis (fig. 27b). Even if the major radial elements are lost by mutation, as in Aaadonta (figs. 28, 29), the microspirals on the apex (fig. 29a) are diagnostic. In a few Hawaiian taxa belonging to Cookeconcha (fig. 30) both macro- and microsculpture are reduced on the apex, while complex postnuclear sculpture is retained. In the Lau Archipelago Priceconcha (Solem, 1973d) only remnants of microsculpture remain on the last two whorls. On early postnuclear whorls the micro- spiral sculpture usually is clearly visible, but on late postnuclear whorls secondary alterations in rib spacing or sculpture reduction frequently obscure this charac- teristic feature. In contrast, the Pacific Island Charopidae have an apical sculpture primitively featuring broadly rounded spiral cords, usually 8 to 12 in number, without any radial elements. Frequently just before the termina- tion of apical growth, one to three low radial undulations will appear in the shell surface. In a few species, the number of spiral cords has been increased to more than 20. and there are 10-20 radial undula- tions on the apical shell surface. Under optical examination, this mimics the appearance of the endodontid apex, but SEM examination shows that the spirals are simple cords and that the radials are surface undulations rather than ribs. Postnuclear sculpture in the Charopidae appears abruptly in a small fraction of a millimeter. Optically it resembles the Endodontidae very closely in having major and minor radials, plus secondary spiral microsculpture. The latter differ in that the microspirals are not FIG. 26. Postapical sculpture of Cookeconcha decussatulus (Pease). West Maui, Hawaiian Islands. FMNH 46605: a, section of body whorl at 300 x showing periostracum flaking from major ribs and difference in shape of periostracal extensions and underlying calcium ridge; b, early postnuclear sculpture at 300 x showing secondary spiral cording, periostracal hairs and flaking periostracum; c, setal area in b at 1,000 x, flaking periostracum at upper right twisted 90 to left and out of normal position. Arrows on a and b point to microspiral riblets in calcium layer of shell. PATTERNS OF MORPHOLOGICAL VARIATION 37 "squiggly," form buttresses to the microradials (for example, Ptychodon microundulata, see Solem, 1970b, pi. 59, figs. 9-11), and thus are much higher next to the riblets than in the trough middle. On late postnuclear whorls the pattern of sculp- ture crowding and secondary modification complicates study of the microspiral elements. But no Pacific Island species has an apical sculpture that departs from the patterns outlined above. In extralimital areas, however, the pattern of apical sculpture is much more complex, with Charopidae from Australia and New Zealand, for example, showing a variety of types (Solem, 1970b, pi. 58, figs. 3-6, for example). Never- theless, in regard to Pacific Island taxa, the nature of the apical spiral sculpture is completely correlated with the anatomically determined family units. It, together with the nature of the apertural barrier microarmature (pp. 63-65; Solem, 1973b), form the only absolute family-level differentiating conchological criteria. Surface wear on the shell is the rule rather than the exception. Frequently the apical whorls will have sculpture remaining only in the sutures (fig. 21c). It is not unusual for specimens collected dead to have nearly all the microsculpture and much of the major sculpture missing. In such situations, cleaning of the umbilicus often will reveal perfectly preserved apical and early postnuclear sculpture. The sculpture is present in both lower and upper parts of the whorls. If transported by a stream of water, the umbilicus may hold an air bubble, while after silting of the opening, the umbilical shell surface is protected from abrasion and acid etching. Even fossils usually have unworn patches of sculpture somewhere on the shell surface or in the umbilicus. Other sculptural elements Additions to the basic sculpture appear macroscopically (lOOx magnification) as one of two types, either secondary spiral grooves as in Mau- todontha zimmermani and all Australdonta taxa (except possibly A. pharcata), or secondary spiral cords in some 50 species. Macroscopic interpretation of the Australdonta sculpture (fig. 124) may be altered when additional SEM studies are made. Photographs of Australdonta raivavaeana (fig. 23) show a tendency towards formation of "spear-point" extensions on the microradials in the supra-sutural region and irregu- larly spaced undulations on the microradials as they traverse the lower whorl surfaces. Whether these are caused by undulations ( = grooves) in the shell surface or by growth changes in the ribbing is unknown. Unquestionably, a multitude of phenomena are gathered under the heading of secondary spiral cording. In species such as Thaumatodon corrugata (fig. 196e), Libera garrettiana (fig. 177c) and Endo- donta binaria (Pfeiffer) the development of low spiral cords is obvious. In others, such as Nesophila capillata (Pease), Libera umbilicata (fig. 178c), Libera bur- satella bursatella (Gould) (fig. 31a), and Nesodiscus FlG. 27. Sculpture of Mautodontha (M.) aoraiensis. Station 863, Mt. Aorai, Tahiti. BPBM 145536 : a, apical sculpture (ca. 300X); b, early postnuclear sculpture, note wavy microspirals (ca. 1,000 X ). taneae (Garrett) (fig. 31c, d), I cannot say whether the protrusions of the riblets result from the addition of very low spiral cords or represent upward protrusions of the riblets themselves. Gambiodonta agakauitaiana (fig. 22d-e) does seem to have low and very broad spiral cords producing the waved effect in the microradial riblets. Aaadonta (figs. 28c-e; 29b) seems to have small knobs on the microradial rib tops that appear as continuous cords under light microscope magnification. Throughout the text all of these phenomena are called "secondary spiral cording." While different structures are involved, reference of more than a few species to any particular type of secondary spiral sculpture is impossible without much additional use of the scanning electron microscope. FIG. 28. Sculpture of Aaadonta constricta constricta (Semper). Station 201, Peleliu, Palau Islands. BPBM 159943: a, entire shell (ca. 50 X ); b, sculpture of apex and postnuclear whorls (ca. 100X ); c, detail of dividing line between apical and postnuclear sculpture (ca. 300x); d, apex and first postnuclear whorl, note spiral "beads" on postnuclear whorl (ca. 300 x); e, suture between apex and first postnuclear whorl, for size compare triangular bit of dirt on apex with d (ca. 1.000X). 38 PATTERNS OF MORPHOLOGICAL VARIATION 39 FIG. 29. Sculptural details on Aaadonta constricta constricta. Station 201, Peleliu, Palau Islands. BPBM 159943: a, spiral cords on apex, note irregular, wavy nature of light cords and broadly waved, lower radial growth wrinkles (ca. 3,000 X); b, detail of postnuclear microradial riblets (ca. 3,000 x). Such secondary sculpture may be present on only part of the shell surface, all of the shell surface, or vary from specimen to specimen of a species. Appear- ance of secondary cording is not random, with only Orangia maituatensis, O. sporadica, and Opanara fosbergi of the 24 Rapan taxa having sculpture. Five of eight Gambiodonta and 8 of 12 Anceyodonta from Mangareva have cording. Eleven of 19 Libera and 5 of 11 Taipidon contrast with only 1 of 15 Minidonta and 2 of 17 Mautodontha having raised secondary spiral sculpture. Some species with secondary spiral sculp- ture are seen in Cookeconcha, Endodonta, Nesophila, Thaumatodon, Pseudolibera, and Kleokyphus. Although 62 of 179 species-level taxa (34.6 per cent) have at least partial development of secondary spiral sculpture, I found no obvious correlations between development of such sculpture and other shell features. The mean adult size of taxa with such sculpture is 4.03 mm. compared with a mean of 4.18 mm. for those without secondary spiral sculpture. The latter category includes a higher percentage of those species lacking major ribbing (23.5 per cent as opposed to 15 per cent for those with secondary spiral cording). Since these are generally the larger species, I suspect the minor size difference is accounted for by this bias and does not signify any significant difference. Possi- bly strong secondary spiral cording in such species as Thaumatodon corrugata and Libera garrettiana is a substitute for lost major radial sculpture. Patterns of variation Variation in the major ribbing involves rib width, rib spacing, and reduction or loss of the sculpture. Most of the smaller species have the individual major ribs narrow, sharply defined and with almost vertical sides that flare outward just before the whorl surface. With increasing shell size, the degree of basal flare is accentuated. Finally, in several very large species such as Gambiodonta grandis and G. agakauitaiana (fig. 22) the major ribs become low swellings on the shell surface. They show scarcely any clear lateral demar- cations. No meaningful measurements of rib width are possible in the latter situation, while the small size and relatively great surface relief of most species defeat any attempt at measuring individual rib widths when they are sharply defined. Considerable effort has been expended in indicating relative rib width in the illustrations prepared under my direction (MM, SH, SG, PR). The Bernice P. Bishop Museum drawings (YK) indicate only rib spacing and give no indication of rib width. Only general discussion of when widening of the ribs has occurred is presented in this report. Adequate quantification proved impossible. Usually this change is correlated with size increase, although the quite small Thaumatodon hystricelloides (Mousson) (fig. 197d-e), T. euaensis (fig. 194a-b), and T. vavauensis (fig. 196a-b) have broadened ribs. Little agreement exists among malacologists concerning how to measure or the utility of rib counts. R. A. Cumber (1960, 1961, 1962, 1964) made consid- erable use of rib counts in analyzing geographic variation in New Zealand charopid land snails. Working primarily with species where there is a sharp break between nuclear and postnuclear sculpture, he has demonstrated least variability in counts obtained from the first or second postnuclear whorl. Successive whorls have noticeably larger standard deviations and standard errors. Most workers have counted ribs on 40 SOLEM: ENDODONTOID LAND SNAILS FIG. 30. Apical microsculpture of: a, Cookeconcha nudus (Ancey). Kawiki, Hawaiian Islands. FMNH 46422. 300 X; b, Cookeconcha (lecussatulus (Pease). West Maui, Hawaiian Islands. FMNH 46605. 300X; c-d, Cookeconcha hystrix (Pfeiffer). Hawaiian Islands. FMNH 46444; c. apex at 300 x ; d, apex-postapical sutural area at 1,000 x. Note periostracal setae on postnuclear sculpture in b. the last complete whorl of growth. Cumber's method reduces intrapopulational variation, but makes a convenient index of rib spacing difficult to calculate. It also depends on having a clearly delineated nuclear- postnuclear boundary. Species of Endodontidae have the apical sculpture continuing onto the postnuclear whorls, with at most a break in spacing that is very difficult to detect. Figures 21a, 27a, and 30a show typical examples where there is no change. Only in Aaadonta (fig. 28c) is a clear change visible. A practical problem of postnuclear whorl-limit recognition exists in the Endodontidae and would prevent use of Cumber's methods. The abundant material available for study was an even greater problem, since counting 60-200 ribs on each shell is quite time consuming. During post-reproductive and gerontic growth there is usually considerable crowding of the major ribbing. Frequently the last few ribs on the body whorl are highly irregular. Rib-count measurements in this study were restricted to adult shells. In the majority of examples there was no clear nuclear-postnuclear whorl bound- ary, either because of growth continuity or surface PATTERNS OF MORPHOLOGICAL VARIATION 41 FlG. 31. Apical and postnuclear sculpture of Libera bursatella bursatella (Gould) and Nesodiscus taneae (Garrett): a, Libera b. bursatella. Station 863, Mt. Aorai, Tahiti. BPBM 142059, sutural area between apex and first postnuclear whorl, note very weak microspiral riblets (ca. 300 x); b-d, Nesodiscus taneae. Borabora, Society Islands. FMNH 91853 ex Fred Button; b, shell apex, note broad radials and very fine and crowded spirals (ca. 300X); c, postnuclear sculpture with reduced major ribbing (ca. 100X); d, detail of postnuclear suture showing rib denticulation of lower whorl (ca. 300 X ). erosion. Rib counts were made on the body whorl (fig. 32a) both to provide a convenient size dimension, shell diameter, for calculating rib spacing and because exact delineation of early postnuclear whorls was impossible. Rib counts were made on about 40-50 per cent of the total adult specimens available. Since the complex apertural denticles found in virtually all Endodontidae provided criteria for species discrimination, detailed analysis of rib-count variation was restricted to a few large samples and the few situations involving subspe- cific taxa. These provide guidelines for interpreting variation in smaller samples. For every species where rib counts could be made, at least 25 per cent of the available adults were checked and species means calculated, provided less than 75 individuals were 42 SOLEM: ENDODONTOID LAND SNAILS FIG. 32. Method of sculptural measurements in the Endodon- tidae: a, raw rib count with shell diameter (A-B) and body whorl sutural diameter (C-D) measurements indicated; b, area on body whorl (A-B) used in determining rib spacing and microriblet counts. involved. If larger numbers were available, about 50 specimens were used. In the Charopidae, where most species lack the apertural denticles, far more attention was paid to shell sculpture variation. Rib-count variation in local populations of Orangia cookei cookei (table LXXXVIII), Rhyso- concha atanuiensis and R. variumbilicata (table LXXXIV), plus Opanara areaensis (table LXXX) indicates only minor variation between populations of the same species or subspecies. The standard errors in these tables are consistently larger than those of Cumber (1960, p. 100) for early postnuclear whorl rib counts comprising the same order of magnitude. This reflects the variation added by grouping shells showing only a short section of gerontic rib crowding with those having up to one-quarter whorl of gerontic growth. While extending the upper range of the normal distribution curve and thus increasing the standard error, such grouping does not seem to have altered the usefulness of these counts. In a sample of 57 Opanara areaensis, for example, the frequency distribution of rib counts is moderately skewed (table XII). The skewness of any such frequency distribution will depend on the mix of barely adult and gerontic shells. Since even the large sample of Libera fratercula rarotongensis showed a slightly skewed normal dis- tribution (figs. 1-4) quite comparable to that cited above, I have proceeded on the assumption that the samples of rib counts are approximately equally skewed unless a lack or surplus of gerontic individuals was noted during measuring. At the level of statistical analysis involved, the error introduced by this variable does not seem objectionable. The major ribs are continuous from the parietal- palatal margin across the periphery, into the umbilicus and up to the columellar-parietal suture. The distance (= shell diameter) from lip edge periphery to the periphery on the opposite side of the shell (fig. 32a, A- B) obviously is greater than from the parietal-palatal suture on one side to the other side (fig. 32a, C-D), yet the rib number is identical. Thus any measure of rib spacing is going to be arbitrary. The ribs are more crowded at the sutures than they are on the periphery, yet they are identical in number and only slightly different in individual rib width at these two extremes. For ease in computation and expression of the zone of major environmental contact, I have chosen an index of rib frequency calculated from the raw rib count and the shell diameter. This is expressed as ribs/mm, on the body whorl. For each specimen this index was calculated by the following formula: Ribs/mm. = rib count on body whorl n x shell diameter in mm. There is an inaccuracy introduced by computing the circumference of a circle where the actual structure is TABLE XII. - RIB COUNTS IN OPANARA AREAENSIS FROM STA. 383 Ribs 45-49 50-54 55-59 60-64 65-69 70-74 Frequency 1 15 9 8 1 PATTERNS OF MORPHOLOGICAL VARIATION 43 one volution of a logarithmic spiral, but the same rib number would hold for both dimensions. Calculation of the growth curve for each specimen or species was impractical, so I have opted cheerfully for simplicity and slight inaccuracy. The bias is equally toward too great an outer whorl distance. An additional in- accuracy results from combining gerontic and pre- gerontic whorl sections with their quite different rib spacings. The frequency distribution for ribs/mm, (table XIII) in the same set of O. areaensis shows TABLE XIII. - PATTERN OF RIB SPACING IN OPANARA AREAENSIS FROM STA. 383 Ribs/mm. Frequency 4.75-5.00 2 5.01-5.25 8 5.26-5.50 16 5.51-5.75 8 5.76-6.00 9 6.01-6.25 8 i 6.26-6.50 4 6.51-6.75 2 even more pronounced skewness, but is typical for measured material. The assumption is made that the same degree and direction of bias exist unless obvious departures were noted either during measurement or initial statistical analysis. Despite the inelegance of this index, it serves to differentiate closely related taxa and provides some data averaging out the number of ribs per unit distance on the shell periphery. The pattern of pregerontic rib spacing is fairly regular, but there are noticeable variations in distance between ribs. Some are caused by injuries, but most have no obvious explanation. As with the Diplommatina studied by Berry (1962), these probably reflect micro- environmental moisture or food fluctuations. For convenience, I have indicated these by stating the approximate width of the interstice compared with the breadth of individual ribs. These observations were taken from the body whorl periphery with the shell positioned for a side view. This area is approximately three-quarters of a whorl behind the aperture and well before the zone of gerontic rib growth (fig. 32b). A visual estimate of "one to two times," or "three to four times" their width was made to encompass variation observed over about one-eighth of the body whorl. * * * * * **.* :* * * * - * * ** FIG. 33. Correlation between rib count and shell diameter in Opanara areaensis areaensis from Station 383. There is also variation in the number of micro- radial riblets between each pair of major ribs. While the microradials often are close to the resolution limit for the stereoscopic binocular microscope at 100 X magnification, they can be counted. Variation in major rib spacing adds to the variation in microrib counts. At the same time that major rib width in relation to the interstices was estimated, I counted the number of microradials present between crowded and more widely spaced ribs on the same portion of the body whorl (fig. 32b). The observed range is cited as "two to r I'M 650 625 I 600 5.25 01 50 55 60 65 Ribs 80 85 FIG. 34. Correlation of ribs and ribs/mm, in Opanara areaensis (all adult material). 44 SOLEM: ENDODONTOID LAND SNAILS TABLE XIV. - CORRELATION OF MICRORADIAL COUNTS WITH MAJOR RIB COUNTS AND SHELL DIAMETER IN THE ENDODONTIDAE Estimated Number of Radial Riblets 1-2 Number of Taxa Median Group Rib Count 250 Mean and Range of Group Rib Count 231.7(195-250) Median Shell Diameter in Group 2.15 Mean and Range of Shell Diameter in Group 2.40(1.97-3.07) 2-4 150 145.1(60.8-225) 3.79 3.62(1.73-5.17) 3-5 19 113 114.1(61.8-250) 3.27 3.33(2.25-6.29) 4-6 27 94 97.8(60.6-178) 3.60 3.64(2.16-6.72) 5-8 36 78 85.9(40.3-202) 3.36 3.56(1.68-6.59) 6-10 18 67 70.2(41.0-108) 3.84 3.85(2.68-5.79) 8-12 25 61 58.9(19-109) 4.02 4.11(2.00-8.46) four" or "five to eight" radial riblets and is based on observation of both specimens with visually crowded major ribs and visually more widely spaced major ribs whenever abundant material was available. The cited figures are estimates and do not represent averaged data. In summary, basic data recorded for each speci- men consisted of raw body whorl rib count and an index of rib spacing calculated from the shell diameter. For each species, based upon observations of a few specimens, estimates were made as to the number of microradials between each pair of major ribs and an estimate of relative rib-interstice width on the first portion of the body whorl. The presence or absence of secondary sculpture was noted and any progressive changes in the postnuclear sculpture during ontogeny were recorded in the written diagnoses. TABLE XV. - SHELL DIAMETER AND RIB SPACING IN THE ENDODONTIDAE Median Ribs/mm. LESS THAN 2 2.00-2.99 3.00-3.99 4.00-4.99 5.00-5.99 6.00-6.99 7.00-7.99 8.00-9.99 10.00-12.99 13.00-19.99 MORE THAN 20 4 9 9 11 8 19 11 20 22 15 5 leU iameter Mean and Range of Shell Diameter 6.40 7.26(4.28-12.26) 4.98 4.91(3.06-7.30) 4.96 5.28(3.88-7.60) 4.51 4.60(2.77-6.59) 4.12 4.03(2.90-4.89) 3.74 3.61(1.87-4.73) 3.36 3.87(2.61-5.42) 3.07 3.32(1.79-6.72) 3.04 2.97(1.73-5.17) 2.84 2.97(1.68-4.61) 2.15 2.49(1.83-3.43) Although there is a general pattern of little difference between populations in respect to rib counts and rib spacing, variation within a particular popu- lation is rather large. Using material of Opanara areaensis as an example, Figures 33 and 34 indicate the typical situation. Rib count does increase with shell diameter (fig. 33), but the correlation (r = 0.56) is not significant. When rib spacing is correlated with raw rib count (fig. 34), the relationship is slightly tighter, but the variation is large enough that individual specimen measurements have little pre- dictive value. The low inter-populational variation reflects the fact that nearly all species of Endodon- tidae are restricted to the leaf litter zone in dense undisturbed forests. Variation in the moisture and temperature environment under these conditions is minimal, hence ecophenotypic changes are predicted to be minor. Correlated variations While statistical treatment of data for individual populations was done in many cases, generally only the mean and range were calculated for each sub- species and species. Although considerable analysis of interspecific sculptural variation patterns has been undertaken, full elucidation of this will be deferred until data from the Charopidae and extralimital taxa are available for comparison. Relationships between shell size, niche and surface sculpture are complex. Only a few of the more obvious factors will be discussed at this time. 1 A. Rib spacing. Within the family Endodon- tidae, 140 species-level taxa retain major radial ribs on the entire body whorl. While in a few of these taxa only very worn or fossil shells were available, for the Data on Priceconcha and Thaumatodon spirrhymatum (Solem. 1973d) are not included below. TABLE XVI. - CORRELATION BETWEEN RIB SPACING AND SHELL DIAMETER IN THE ENDODONTIDAE Ribs /mm. X Diameter ca H co CO 0) ON ON CM O O CM ON ON 1 O o ON .l 15.0 85.0 36.1* 63.6 61.8 38.2 93.7 6.3 98.0 2.0 21.1 71.0 7.9 1.3 8.2 77.7 11.14 75.0 21.. 3 87.5 12.5 28.6 71.1* 90.9 9.1 73.2 22.1 60.0 1.0.0 75.0 25.0 25.0 75.0 25.0 75.0 15.0 85.0 38.1 61.9 1.3 FIG. 42. Palatal barrier sculpture in Hawaiian Endodontidae: a-d, Endodonta fricki (Pfeiffer). Waianae Mts., Oahu, Hawaii. BPBM 128063. a, lateral view of 2nd palatal at 2,250 X; ft, top view of 2nd palatal at ll.OOOx; c, view from aperture of 2nd palatal lamellar sculpture at 2.750X ; d. same at 6.800X ; e-f, Coohecimchti nudus (Ancey). Kaiwicki, 2,500 ft. elevation, Hilo, Hawaii. FMNH 90319. e, top view of 2nd palatal at l.OOOx ; f, same at 5,000 x . Photographs courtesy of Engis Equipment Company, Morton Grove, Illinois. 66 PATTERNS OF MORPHOLOGICAL VARIATION TABLE XL. - PERCENTAGE DISTRIBUTION OF PALATAL BARRIER NUMBERS 67 Species Mautodontha parvidens M. rarotongensis M. boraborensis Opanara bitridentata Rhysoconcha atanuiensis R_. variumbilicata Ruatara p_. oparica R_. o_. normal is Orangia sporadica Australdonta degagei A. pseudplanulata Taipidon woapoensis Libera b_. bursatella L_. b_. orofenensis L_. spur i a L_. garrettiana L_. heynemanni L_. incognata L_. jacquinoti L_. subcavernula L. tumuloides Number of palatal barriers 123 5.1 6.3 Thaumatodon multilamellata Australdonta degagei, Ruatara, Rhysoconcha, and Libera bursatella are based on quite significant specimen numbers. The low frequency deviants pre- sumably thus reflect rare mutants or developmental accidents. Other situations, as in the Australdonta, hint at classic Mendelian dominance as the underlying basis of variation. In discussing the correlatives of barrier numbers, the varying extent of departures from a uniform number presented practical difficulties. Species were classed as having significantly variable barrier counts only if one-third or more departed from the standard number. Otherwise the species is grouped with those taxa having the predominant character state condition. While this underemphasizes the total extent of variation, it does facilitate making comparisons. Most genera have a clear majority or nearly all species showing a constant number of parietal barriers (see table XLI). Only in Mautodontha, Anceyodonta, 66.7 7.0 33.3 5.3 5.1 25.0 89.8 62. U 3.2 7-9 19.2 25.0 12.0 6.3 9-7 k.6 86.7 11.8 3.8 It 86.0 88.2 78.1 86.3 73.6 82.0 3.5 75.0 88.0 87.1 95^ 13.3 23.5 96.2 5 7-0 11.8 15.1 98.0 96.0 13.0 93.6 9^.8 95-3 5.2 6.1 2.0 U.O 5.0 66.7 33.3 Australdonta, and Taipidon are there massive ex- perimentations in barrier numbers. Genera such as Minidonta, Cookeconcha, Nesodiscus, Gambiodonta, and Thaumatodon show some variations. The genera with strong variations are also those in which there is a great range in the degree of apertural narrowing achieved by the barriers (table XLVIII). Both size and number reduction of the barriers are involved in this pattern. Genera with a predominance of 2 parietals include Cookeconcha and Endodonta from Hawaii, Orangia, Taipidon, and Libera. Minidonta, Australdonta, Opa- nara, and Aaadonta have mostly 3 parietals, while Thaumatodon has 4 parietals, and the Mangarevan Anceyodonta and Gambiodonta tend to have 4 or 5 barriers. All the taxa with many parietals involve extreme cases of size reduction for the barriers. Thus the number of parietal barriers is tied to phyletic units and also is involved in the degree of apertural narrowing. Raw data on the size correlations of SOLEM: ENDODONTOID LAND SNAILS TABLE XLI. - PHYLETIC CORRELATION OF PARIETAL BARRIER NUMBERS Minidonta Mautodontha Anceyodonta Cookeconcha Kleokyphus Opanara Rhysoconcha Ruatara Orangia Australdonta Taipidon Planudonta Rikitea Nesodiscus Nesophila Kondoconcha Endodonta Pseudolibera Li"bera Gambiodonta Thaumatodon Zyzzyxdonta Aaadonta Priceconcha TOTALS 1, 1-2 1 2 Number of parietals 2 2-3 3 3-U ^ 2 6 12 5 1 l 1 5 1 3 1 1 2 1 8 IT l many 12 3 2 1 8 1 3 5 2 1 2 1 5 U l l l 2 1 1 2 1 8 3 6 19 25 1 8 parietal barrier number (table XLII) require inter- pretation. The larger size of those taxa with only 1 parietal reflects the concentration of this state in species of Cookeconcha and Nesodiscus in which the apertural narrowing function has, for all practical purposes, been lost. Similarly, the larger size of those taxa with 2 parietals results from the phyletic units Endodonta and Libera clustering there. The large standard errors of the mean show that there are not significant size differences between barrier numbers, but that this relates more to phyletic unit and degree of apertural narrowing. The number of palatal barriers is basically 4 in the Endodontidae (table XLIII), with changes to 3, 5 or total loss of palatal barriers the most common alterations. Addition of a 5th palatal has happened in several different ways. Most Aaadonta, Kleokyphus callimus (fig. 95b), Endodonta lamellosa, and E. marsupialis (fig. 167a) have an extra subpe- ripheral barrier; in Australdonta (fig. 127a, b) it is possible that descent of the former columellar barrier onto the lower palatal wall has been completed; Rhysoconcha (fig. 112) and Minidonta micraconica (fig. 65b) have extra supraperipheral barriers; while PATTERNS OF MORPHOLOGICAL VARIATION 69 TABLE XLII. - SIZE CORRELATION OF PARIETAL BARRIER NUMBERS TABLE XLIII. - PHYLETIC CORRELATION OF PALATAL BARRIER NUMBERS Number of parietal Number barriers of taxa X D 7.146 X W 7+ 1, or 1-2 19 5.310.11 5.9^0.23 (2.90-11.19) (k 3/8-8) 6k U. 66*0.20 5.880.13 (1.79-8.99) (3 5/8-8) 2 or 3 k 3.630.29 5.UO+0.18 (3. 01-!*. 33) (5-5 3/U) 3 U 9 3.220.13 5.330.09 (1.68-6.60) (U-8) 3 or U It 5 3.^1*0.25 5.690.25 22 3.81+0. U6 5.760.19 (2.00-12.26) (U-7 1/2) 1* or 5 2 3.080.15 6.00+0.1*3 ^ (2.93-3.23) (5 1/2-6 1/2) 5 7 3.1(00.12 6.030.19 (2.20-5.07) (5 3/8-6 3A) many threadlike 6 !.981.28 5.>*00.17 (3.21-11.29) (1* 3A-5 7/8) Kondoconcha othnius (fig. 162b) has both enlarged the upper palatal and added another supraperipheral palatal barrier. Reduction to only 3 palatal barriers generally involves loss of the supraperipheral barrier, as in Minidonta inexpectans (fig. 62d) and some Aaadonta (fig. 206b), or by elimination of a subpe- ripheral barrier, as in many Libera (figs. 173d). Unlike the number of parietal barriers, there is a rather clear size correlation with the number of palatal barriers (table XLIV). Although the large standard errors of the mean for the diameters indicate a large amount of variability, the trend itself is obvious. Table XLV reviews the correlations between numbers of parietal and palatal barriers. There is generally central clustering. The pattern of reductions in both sets has been described above. Barrier length is a difficult measurement to make and one that is subject to considerable error, since the distance must be judged by looking into the aperture at an angle. In addition, the growth pattern outlined above would mean that the length of barriers would vary according to the stage in the growth cycle at which the animal died. Given the above factors, the cited lengths should be taken for what they are, estimates of relative length and not firm figures. Figure 43 shows how length was estimated, as an arc of a circle to the nearest sixteenth. The pattern of the parietals, normally extending anteriorly of the lip edge, permits seeing slightly more than one-fourth of a whorl from its anterior termination. If the sharp posterior descension could be spotted, but not the actual posterior margin, it was scored as "to the line of Minidonta Mautodontha Anceyodonta Cookeconcha Number of palatals 1, 1-2, 2 3 3-5 It 1 I many Orangia Australdonta Taipidon Planudonta Rikitea Hesophila Nesodiscus Kondoconcha Endodonta Pseudolibera Libera 3 9 2 10 2 12 3 1 It 2 1 1 9 1 2 3 5 1 1 2 5 1 1 7 1 1 Gambiodonta Thaumatodon Zyzzyxdonta Aaadonta Priceconcha TOTALS 5 -5 6 22 13 1 2 23 It 1 25 vision" or "one-quarter whorl," depending on the degree of anterior extension. Extending the same principle to measuring the palatal barrier length was simple, but this means that comparisons of parietal and palatal lengths is not possible, since the two arcs are at different distances from the shell axis. In Figure 43, the solid lines indicate the arc for a one-quarter- whorl long parietal barrier, while the dotted lines indicate the length of a palatal barrier, in this case about one-eighth whorl long. TABLE XLIV. - SIZE CORRELATION OF PALATAL BARRIER NUMBERS Number of palatal Number barriers of taxa X D X W 21 5.030.52 5.280.l8 (2.90-11.19) (l*-7) 1-2 lit 5.680.60 6.500.28 (1.97-11.29) (U 3A-8) 3 22 1*.990.1*7 6.000.2l( (1.68-12.26) (l-7 5/8) 3-5 !*.l60.35 5.510.ll* (2.70-5.17) (5-6 lA) It 85 3.710.ll* 5.650.08 (1.73-8.99) (3 5/8-7 5/8) 5 2k 3.510.26 5.>60.ll* (1.97-7.20) (it lA-7 1/2) many k 3.760.3>t 5-98+0.10 (t. 13-5. 77) (5 3A-6 lA) TABLE XLV. - CORRELATION OF PARIETAL AND PALATAL BARRIER NUMBERS Number of parietals Number of palatals 1-2 2-3 many, a few high many, none high no data 1-2 2-3 3-5 5-6 many 1 8 U 3 1 2 1 2 5 1 1* 1 12 3 28 6 2 1 3 1 1 6 2 30 9 3 2 1 2 13 6 1 2 7 1 2 3 1 TABLE XLVI. - PARIETAL BARRIER LENGTH AND SIZE CORRELATIONS Length of parietal barriers Number of taxa X diameter X whorl count l/8th whorl 1 3.88 5 .25 less than 3/l6ths 11 3. It 6*0. It li 5 .28+0.37 3/l6ths 31* 3.7810.25 5 .U6+0.17 less than 1/Uth 21 3.9110.29 5 .7310. lit 1/ltth 33 3.5210.22 5 .18+0.11 more than 1/ltth 18 3.260.l6 5 ,5710. lit to line of vision 16 It. 14*0.28 6, .7810.18 beyond line of vision Ult lt.92tO.3li 6, ,0610.13 TABLE XLVII. - PALATAL BARRIER LENGTH AND SIZE CORRELATIONS FIG. 43. Method of measuring parietal (solid linel and palatal (dotted line) barrier lengths. Length of palatal barriers Number of taxa X diameter X whorl count less than l/8th 13 3.614+0.29 5.l60.l8 l/8th 69 >4.11*10.22 5.7710.11 3/l6ths 38 U.100.2li 5.85*0. Ill 1/ltth 18 3.570.21 5.7110.12 line of vision It 3.2110.U2 6.0liio.l9 beyond line of vision 9 3.870.58 6.0510.36 70 PATTERNS OF MORPHOLOGICAL VARIATION TABLE XLVIII. - DEGREE OF APERTURAL NARROWING BY BARRIERS 71 Genus Minidonta Mautodontha Anceyodonta Cookeconcha KLeokyphus Opanara Rhysoconcha Ruatara Orangia Australdonta Taipidon Planudonta Rikitea Mesophila Nesodiscus Kondoconcha Endodonta Pseudolibera Libera Gambiodonta Thaumatodon Zyzzyxdonta Aaadonta Priceconcha Total taxa 15 IT 12 16 2 12 2 h 5 12 11 k 1 2 10 1 8 1 19 7 9 1 9 1 181 Strong 2 8 (1-) 2 (2-) 2 1 (IH 1 1 U-) 1 (1-) 6 7 2 1 2 (1-) 36 (7 ) Apertural narrowing Moderate 8 5 (-1) U 3 2 8 (2-) 1 1 (1-) U (1-) 5 (5-) (-1) 7 (1-) Weak 5 (2-) 6 (-2) 1 1 3 1 (-1)1 1 (-1)10 (1-*) (-3) 7 HO 7 (1-*) HO 7 J^ ( 12)7U (12 ) ( 5)30 (2 ) Not 6 7 It 2 U 2 10 1 2 There is no major correlation of parietal barrier length and size (table XLVI) until the very elongate barriers are encountered. Those that extend to or beyond the line of vision are found in species with higher whorl counts (and thus larger size). Most probably this is a simple release pattern. Whatever triggered the adding of additional whorls also slowed the posterior resorption phase of barrier growth, or the barriers simply lengthened proportionately to the increment in whorls. Palatal barrier length correlates probably more with palatal barrier shape changes (table XXXII) than it does with size (table XLVII). Short barriers are a prelude to loss of the barriers. Degree of apertural narrowing The most fundamental functional significance of barriers is the degree to which they effectively narrow the aperture. Ideally this should have been quantified in some manner. An obvious procedure would have been to measure the cross-sectional area of the aperture in a plane through the expanded portions of the barriers, subtract from this the area that is open between the barrier tips, and then calculate the per cent of the aperture that is walled off by the barriers. Unfortunately, this proved to be impractical, since sufficient material to permit sectioning each species through the barriers was not available. In many species, such as Libera (fig. 182) and Gambiodonta 72 SOLEM: ENDODONTOID LAND SNAILS (figs. 187a, c), the elevated portions of the barriers are recessed essentially beyond the line of vision from the aperture. Thus even direct visual estimates of the degree to which the apertures are narrowed become subjective. The degree of narrowing obviously varies widely. In all Nesodiscus and taxa with reduced barriers, such as many Mautodontha (fig. 78) and Cookeconcha (fig. 94), the aperture is virtually unimpeded by the barriers. In contrast, taxa such as most Anceyodonta (figs. 81, 82, 83) and Endodonta (fig. 167) have the aperture almost closed by the thicket of protruding barriers. An estimate of relative closure is tallied in Table XLVIII. The number of taxa that are intermediate in character have been indicated in parentheses and are accompanied by directional arrows. Genera such as Anceyodonta and Gambiodonta share common an- cestry, with the latter retaining strong apertural constriction despite their large size, as do the Hawaiian ground-dwelling Endodonta, although many Cook- econcha that are almost equivalent in size have almost completely lost their barriers. Where a variety of relative constrictions is found in a genus, such as Minidonta, Mautodontha, Cookeconcha, and Taipi- don, there has been drastic size reduction of the barriers (table XXXVII). As an indication of the different states, reference is made to species of Minidonta, which range from strongly restricted to nearly no major narrowing. Minidonta micro (fig. 63a) is strongly narrowed; M. micraconica (fig. 65b), moderately; M. hendersoni (fig. 63c), weakly; and M. planulata (fig. 69b) tends towards virtually no effec- tive apertural narrowing, a state achieved in such species as Mautodontha subtilis (fig. 77d). At least one additional factor is involved, the body whorl contour (fig. 14A-E). Species with the body whorl laterally compressed (fig. 14A), such as Mau- todontha maupiensis (fig. 76a), will have the aperture more constricted than species with an evenly rounded periphery, such as Mautodontha subtilis (fig. 77d), even though the actual height of the barriers may be virtually identical in the two species. The probable extent of this correlation is shown in Table XLIX. TABLE XLIX. - CORRELATION OF APERTURAL NARROWING AND BODY WHORL CONTOUR Body whorl Compressed laterally Evenly rounded Strong 6 Ik Apertural narrowing Moderate Weak 19 li 15 23 12 5 12 8 Taxa with evenly rounded peripheries tend to have much less apertural narrowing than do those taxa with either laterally compressed or keeled peripheries. Formation of a keel would automatically tend to accentuate the degree of apertural narrowing by the barriers, as would the lateral compression of the body whorl. But the latter change may be directly selected for because of the aperture narrowing, while keel formation (pp. 21-23) may be more of a means of lessening shell height, with the apertural narrowing a secondary feature. Some of the largest (Endodonta, Gambiodonta) and some of the smallest (Minidonta, Cookeconcha) taxa have strongly narrowed apertures, so data on overall size are not presented. Often, within a particular genus, species with strongly constricted apertures will be smaller in adult size than those with little or no apertural narrowing. This is obvious in Anceyodonta, where the eight species with strongly constricted apertures have an average mean adult diameter of 2.54 mm., while the four species with only moderately constricted apertures have an average mean diameter of 3.42 mm. Taipidon petricola petri- cola, the only species in that genus with a strongly constricted aperture, is the smallest in the genus. Similarly, the shift from "moderately" to "weakly" to "not constricted" in Australdonta is size correlated, the respective average mean diameters being 3.38 mm., 3.95 mm., and 4.19 mm. Thus within a lineage there may be a tendency for barrier constriction reduction to correlate with size increase, but there is no overall pattern. Summary of barrier variation Much of the variation in structure, reduction, and numbers of apertural barriers in the Endodontidae is correlated with phyletic factors rather than size. There is no equivalent in respect to the apertural barriers to the apparent triggering size of 4.75 mm. after which drastic radial rib reduction occurs (pp. 47-49). The degree of superior expansion to the barriers does correlate partly with small size, but lack of any expansion is a correlative of barrier reduction rather than of shell size. Relative expansion of the parietal and palatal barriers is not closely correlated. Recession of the palatal barriers does correlate with size change, but shape and form of the barriers correlate with each other, not shell size. Similarly, there are correlations in form and position of the columellar barriers, but these are independent of shell size. Within phyletic lines there are tendencies toward barrier size reduction and lessened degree of apertural constriction, both of which correlate to some extent with increased size within that lineage. But this varies from lineage to lineage in terms of what size is reached before major barrier reduction occurs. As a result of the above data, I have confidence that apertural barriers carry a high information content for phyletic deductions. GROSS ANATOMY This monograph covers 185 species-level taxa. No material was seen of six previously described species. The dissections include 58 taxa, whose phyletic PATTERNS OF MORPHOLOGICAL VARIATION 73 TABLE L. - PHYLETIC REPRESENTATION OF DISSECTED TAXA Genus Minidonta Mautodontha Anceyodonta Cookeconcha IQeokyphus Opanara Rhysoconcha Ruatara Orangia Australdonta Taipidon Planudonta Rikitea Nesodiscus Nesophila Kondcconcha Endodonta Pseudolibera Libera Gambiodonta Thaumatodor. Zyzzyy.dorvto. Friceconcha Aaador.ta Totals Total taxa 15 17 12 18 2 12 2 k 5 12 11 ll 1 1 10 1 19 7 9 1 1 _9 185 Number seen 15 17 12 16 2 12 2 1* 5 12 11 It 1 8 2 1 8 1 19 7 9 1 1 _9 179 Number dissected 1 2 11 2 2 1* 2 6 3 2 1 3 6 58 distribution is summarized in Table L. Partial speci- mens of Opanara areaensis microtorma, Orangia cookei tautautuensis, and Kondoconcha othnius were also seen, but the material was too fragmentary or poorly preserved for successful illustrations to be prepared or measurements made. The major phyletic gaps in the coverage are in the Mangarevan radiation (many Minidonta, and all Anceyodonta, Gambio- donta, and Rikitea}, Minidonta, Mautodontha, and Kleokyphus. Cooke (1935, pp. 41-42) reported that the complete destruction of native forest on Mangareva had occurred prior to 1934, although obviously live collected native land snails from the island were added to museum collections as late as 1872. Similarly, the collections on the Tuamotus have yielded virtually no living endemics in this century. The faunas of these two areas include Pseudolibera, both Kleokyphus, Mautodontha daedalea, five species of Minidonta, all Anceyodonta, all Gambiodonta, and Rikitea insolens, for a total of 30 species-level taxa. If these groups are eliminated, then 38.9 per cent (58 of 149) species-level taxa were dissected at least in part. This section includes data on Thaumatodon spirrhymatum and Priceconcha tuvuthaensis, the species described in Solem (1973d). Because of the fragmentary material available for dissection in many species, information concerning apical organ systems frequently is lacking. For ex- ample, the origin of the penial retractor muscle could be observed in only 40 of the 58, the length of the pallial cavity in 35, and the character of the ovotestis in 37 taxa. Penial length was measurable, however, in 54 of the 58 taxa. Within these limitations of materials and successful observations, the following pages review the observed patterns of variation in structure, delineate the topographic anatomy of the body, compare and contrast these features with equivalent structures as observed in the Pacific Island Charo- pidae, and occasionally add comparative remarks from dissections of Australian, New Zealand, New Caledo- nian, and Lord Howe Island taxa. These additions are background data necessary to establish the importance assigned variations in structure against the context of a broader scope investigation. While comparisons of the endodontid shell variation with the Charopidae have been deferred until the second monograph, the anatomical data are presented here. The sequence of structures places the genital system first because of its higher information content concerning phylogeny within the family. GENITAL SYSTEM The terminology used below is modified from that of H. B. Baker (1938b, pp. 6-10, 92), since it more adequately reflects the apparent functioning of the system than does the terminology used by English workers such as Rigby (1963, 1965). The latter system was developed by forced comparisons of prosobranch and very advanced pulmonate taxa. It is not transfer- able to description of less specialized pulmonate taxa. Many of the fusions and specializations found in the advanced taxa have not taken place in the more generalized families. A recent commentary (Bayne, 1973) also basically adopts the Baker terminology. The abbreviation used in the illustrations follows each term. OVOTESTIS (G) - The hermaphroditic gland or ovotestis is located above the reflexion of the intestine from the stomach apex. Typically it consists in the Endodontidae of many multilobate alveoli strung along a single collecting tubule. Typical patterns are shown in Figure 164a for the sequential relations of the alveoli along the collecting tubule and Figure 165e for the palmately clavate branching pattern of a single unit. The collecting tubule runs along the lower parietal wall of the whorl cross-section, ascending for one-third to two-thirds of a whorl, depending on the length of the ovotestis. I have no data on seasonal variation in development of individual follicles or length of the entire organ. Preservation in this area of the animal generally was poor. This dictated drawing only that portion which could be extracted easily, so that apparent differences in ovotestis lengths shown in the drawings are artifacts of dissection resulting from preservation problems. In the Endodontidae there is a single characteristic pattern of ovotestis orientation. 74 SOLEM: ENDODONTOID LAND SNAILS FIG. 44. Ovotestis structure and position in the Endodontidae (a- 6) and Pacific Island Charopidae (c-e): a, typical pattern as seen in Endodonta fricki; b, pattern after nuclear whorl enlargement as seen in Nesophila tiara; c, typical pattern in the Pacific Island Charopidae as seen in the Tongan species usually known as "Charopa" vicaria (Mousson, 1871); d, pattern seen in Aeschro- domus stipulata (Reeve, 1852) from Pelorus Bridge, Marlborough, New Zealand (FMNH 165395), probably correlated with changes in coiling pattern and whorl size; e, pattern seen in Thalassohelix propinqua (Hutton, 1883) from Weka Pass, Waikari, South Can- terbury, New Zealand (FMNH 165399) and associated with size increase of the whorl cross-section. (CK) This is shown in Figure 44a. The follicles extend upward and outward at an acute angle to the shell axis. Often they reach from the lower parietal to the upper palatal margin, being imbedded in digestive gland tissue. Up to a full whorl or whorl and a half of digestive gland tissue extends above the ovotestis apex. This is modified in only three of the dissected taxa. In Nesophila tiara (Mighels) and Cookeconcha jugosus (Mighels) (figs. 165d, h; 44b) the follicles have a right angle radial orientation to the shell axis, instead of sitting at an acute angle, while in both species of Rhysoconcha (p. 255) the follicles are reduced in number and lie essentially parallel to the plane of coiling. Cookeconcha jugosus is 25 per cent larger in diameter than an apparently related species, C. hystricellus (Pfeiffer), with virtually identical whorl count. The latter species (fig. 165j) retains the normal positioning of the ovotestis. Nesophila tiara has a mean shell diameter of 11.29 mm. with an average of slightly less that 5V& whorls, while the other studied Nesophila, N. capillata (Pease), has a mean diameter of 4.46 mm., with slightly less than 4% whorls. Visual inspection of these four species shows that the apical whorls of the pair with altered ovotestis orientation are significantly larger than those of their smaller relative. The ovotestis orientation in N. capillata is unknown. The shell size increase in these taxa resulted not from the addition of more whorls, as is the pattern in Nesodiscus and Libera, but from proportionate change in whorl size without increase in whorl number. Thus the shift of ovotestis orientation in the Nesophila and Cookeconcha probably reflects the sudden availability of extra space in the upper whorls, and not any shift in genital structure itself. As presented elsewhere (pp. 255-256), the change in follicle orientation seen in Rhysoconcha is one of the features that led me to propose that this is an example of secondary size reduction. The Pacific Island Charopidae have ovotestis patterns that stand in great contrast to those found in the Endodontidae. The typical charopid pattern (fig. 44c) is for one or two clusters of palmately clavate alveoli to extend virtually horizontally from the stomach apex toward the apex of the viscera. A single clump will extend as shown; the second clump, when present, will lie apicad of the first, separated from it by a distinct zone of digestive tissue. In such taxa as Aeschrodomus stipulata (Reeve, 1852) (fig. 44d) from New Zealand, which have an altered shell shape with high spire, the two clumps may nearly fill the whorl. Only the narrowest strips of digestive gland tissue extend anteriorly. Multiple experiments in size in- crease have occurred in the Charopidae, with forms such as the New Zealand Thalassohelix propinqua (Hutton, 1883) (fig. 44e) showing nearly right angle to the shell axis orientation for the follicles. The basic two-clump format is clearly visible, despite the convergence to the endodontid patterns (fig. 44a, b). In regard to Micronesian, Polynesian, and Melanesian PATTERNS OF MORPHOLOGICAL VARIATION 75 taxa, the difference in ovotestis structure is diagnostic at the family level. The Charopidae not only have fewer and proportionately larger alveoli in the ovotestis, but they are oriented and clumped differently from the follicles found in the Endodontidae. Observed changes in the follicles seen in the Endodontidae correlate with major shifts in growth pattern, either increase in whorl cross- sectional area (Nesophila and Cookeconcha jugosus) or probable secondary size reduction (Rhysoconcha). HERMAPHRODITIC DUCT (GD) - This is an expanded continuation of the joined collecting tubules of the ovotestis follicles that extends to the carrefour region. In nearly all generalized shell-bearing pulmo- nates, this tube extends along the parietal or lower parietal margin of the whorl beneath or adjacent to the stomach. The function of this tube is to transport sex products. Its existence as a separate organ may be the result of simple space problems. The stomach occupies the same part of the visceral hump as the hermaphroditic duct, the space-consuming ovotestis lies above the stomach expansion and initial intestine reflexion, while the large albumen gland occupies the area below the stomach. The hermaphroditic duct provides an essential narrow passageway through the zone occupied by the main expansion of the digestive tract. In 32 of the examined Endodontidae, the hermaphroditic duct is a simple expanded tube (fig. 45a). The outer edge may have a few humps or wrinkles in it, but the inner surface is smoothly and evenly curved. Inspection of the drawings scattered through the systematic section shows major differ- ences in degree of expansion and the extent to which it is obviously reflexed anteriorly before joining the carrefour. The expansion differences probably vary with the state of reproductive activity, while the exact degree of reflexion more probably reflects the state of contraction of the animal when preserved. The existence of a slight reflexion in the talon area is normal, but this often is accentuated in contracted animals. Presumably this is caused by apical move- ment of the pallial apex forcing the albumen gland and intestinal loops slightly above the stomach expansion by compressing the early part of the stomach. The permanent bend in the hermaphroditic duct permits varying lengths to be reflexed upward. Artificial straightening of the reflexed section can happen quite easily while pinning out a dissection for drawing. Thus apparent changes in this section of the genital system as illustrated have no significance. What does seem to have significance is the degree to which the duct is kinked or partly coiled. This is shown most completely by Aaadonta kinlochi (fig. 45d) and only slightly less so in Ruatara oparica normalis (fig. 64h) and R. o. reductidenta. In Cooke- concha jugosus (fig. 165h) the middle half of the duct is kinked, while in Thaumatodon euaensis (fig. 195c) and Aaadonta fuscozonata fuscozonata (fig. 199e) the FIG. 45. Hermaphroditic duct variation in the Endodontidae: a, typical pattern; b, slight kinking as in Thaumatodon hystricelloides; c, moderate kinking as in Aaadonta constricta; d, extensive kinking as in Aaadonta kinlochi. In each drawing the ovotestis would be on the left and the reflexion to the carrefour on the right. (CK) upper half of the duct is kinked. In Thaumatodon hystricelloides (fig. 45b) the top quarter is coiled, as is the top third in Aaadonta c. constricta (fig. 45c). The differences between the several states are shown and suggest that a progressive pattern of kinking from the apex is normal. The normal pattern in the Charopidae from the Pacific Islands also is for a simple her- maphroditic duct, but coiling occurs in many Austrozelandic taxa, such as Allodiscus dimorphus (Pfeiffer, 1853). Generally the duct in the Charopidae is thicker, shorter (corresponding with a general tendency toward fewer whorls in that family), and appears more highly iridescent than those from endodontid species. Viewed within the context of their lineages, the few endodontid taxa with coiled or kinked hermaphro- ditic ducts show no clear shell correlations with each other. Kinking has occurred in flat-spired, large species (A. kinlochi and C. jugosus), relatively small species with spires that protrude much higher than normal 04. c. constricta, A. f. fuscozonata, T. euaensis, and R. o. reductidenta) and species with a median spire projection (T. hystricelloides and R. o. normalis). The phenomenon is not associated with whorl increase and high spire, as in Libera and Nesodiscus, but with spire protrusion while retaining a normal, or even slightly lowered whorl count (A. kinlochi, T. euaensis, A. c. constricta). In the Austrozelandic Charopidae, kinking is, in general, associated with visceral hump reduction, changing in coiling patterns of the shell, and major size increments. Thus kinking of the hermaphroditic duct in the Endodontidae should be viewed as having occurred independently in each genus. The presence of a kinked duct in an endodontoid genus should be taken as a signal that changes have occurred from the visceral 76 SOLEM: ENDODONTOID LAND SNAILS hump coiling or protrusion pattern found in its ancestors. The reasons for kinking rather than simple shortening of the duct is unknown. There could be functional significance requiring a minimum duct length. It could be merely an accident of development, with the trigger that results in shortening the stomach failing to influence the development of the duct itself. Although no measurements were taken as docu- mentation, simple observation confirms that the longer the stomach, the longer the hermaphroditic duct, since it serves to convey sex products from a large organ above the stomach to a large complex of organs below the stomach. TALON (GT) and CARREFOUR (X) - Consid- erable controversy exists concerning the structure and function of organs in this area. Many English workers refer to this as the "receptaculum complex" (Rigby, 1965, pp. 455-457), have called the lower portion of the hermaphroditic duct a "seminal vesicle," and inferred that there are functional divisions into a "fertilization sac" and two "receptacula seminis" in Succinea and zonitids such as Oxychilus (Rigby, 1963, pp. 328-329, fig. 7). In the Bulimulidae, Van Mol (1972, pp. 199-202) reported a highly complex talon structure, part of which is a fertilization chamber. In other taxa, such as the Australian camaenid genus Craterodiscus, there is no differentiated structure in the area where the hermaphroditic duct joins the albumen gland, prostate, and uterus (Solem, 1973c, p. 379, fig. 1, d). Similarly, in the Tornatellinidae (Cooke and Kondo, 1960, pp. 31-33) the structures can be greatly reduced (as for example in Tekoulina, Solem, 1972b, p. 101, fig. 3, b) or with an accessory, very large organ (Cooke and Kondo, 1960, pp. 31-33, fig. 8). In contrast to this, both the Endodontidae and the Charopidae have a relative- ly simple structural arrangement. The talon (GT) is a bulbous expansion sitting on a shaft that may be extremely short (most Charopidae) to usually quite long (many Endodontidae) and that emerges from the carrefour (X) apex. There is no external indication of any subdivision to the talon. In several exceptionally well-preserved specimens it was possible to examine this region in some detail. The Endodontidae have variation in the talon shaft length. Of the 41 taxa in which this could be observed, 25 had a long shaft, as in Endodonta fricki (fig. 164b); seven had a medium length talon shaft, as in Aaadonta c. constricta (fig. 199c); and nine had a short talon shaft, as in Thaumatodon euaensis (fig. 195c). These variations are only partly phyletically associated. Of the genera in which two or more species could be scored, Cookeconcha, Rhysoconcha, Orangia, Australdonta, and Libera had only long talon shafts. Both Planudonta in which the talon could be observed had short shafts, but the four Taipidon had two with short and two with long shafts. Opanara has species showing all three states, while Aaadonta, Thaumato- don, and Taipidon have species with two states. The situations involving mixed shaft types within genera or involving derivative genera (Taipidon and Planudonta) require special examination. In Opanara, short shafts are possessed by O. bitridentata (fig. 96c) and O. duplicidentata (fig. 96g); medium-length shafts by two subspecies of O. areaensis (fig. 96j); and long shafts by O. altiapica (fig. 97a) and O. megomphala tepiahuensis (fig. 97e). These are the only species in that genus for which this character was observed. O. areaensis microtorma, O. duplicidentata, and O. bitridentata are sympatric at Stations 446 and 451 on Mt. Perahu, while the nominate race of O. megom- phala occurs with O. areaensis at Station 477. O. duplicidentata and O. bitridentata differ markedly in penial length (1.5 and 2.65 mm. long, respectively), pilaster pattern (figs. 96d, h), and origin of the columellar retractor muscle. O. areaensis has the penis length and columellar muscle origin of O. bitridentata, but modified pilasters (fig. 96k). The other Opanara mentioned above also differ in penis size and pilaster patterns. Of the Marquesan species, Taipidon semi- marsupialis (fig. 139a), T. fragila (fig. 138e), Plan- udonta intermedia (fig. 147a) and P. concava (fig. 147c) have short talon shafts, while T. petricola petricola (fig. 138a), T. p. decora (fig. 49a), and T. centadentata (fig. 139e) have long talon shafts. Both Planudonta, T. semimarsupialis, and T. centadentata are sympatric on Mt. Ooumu, Nukuhiva Island. Taipidon fragila from Hivaoa essentially is sympatric with T. varidentata, a species whose talon length is unknown, while the two races of T. petricola are the only endodontids known from Eiao and Hatutu Islands. The several Thaumatodon and Priceconcha are all allopatric, but they have short, medium, or long talons. In Aaadonta constricta and A. fuscozonata, which were collected together at Station 203 on Peleliu and Station 221 on Koror, there are apparent differences in talon length. Because of the allopatric variability in Thaumato- don the evidence is not fully conclusive, but there is a strong correlation of talon shaft length variations occurring under conditions of congeneric sympatry. This suggests that this feature may be involved in species isolating phenomena. Much more study is needed before this can be stated as a fact or disproved. Certainly the variations are repeated in several phyletic units, so that selection on a relatively low level would be involved. Because of its numerical frequency and broader geographic distribution, I conclude that the long talon shaft is the generalized state in the Endodontidae. The carrefour (X) is a barely noticeable to prominent swelling in the gonoduct that usually lies partly buried in the surface of the albumen gland (GG). It receives ducts from the hermaphroditic duct (GD), talon (GT), and albumen gland (GG). Separate PATTERNS OF MORPHOLOGICAL VARIATION 77 ducts exit to the uterus (UT) and prostate (DG). In several taxa, preservation was excellent and the apparent relationship of these ducts could be observed using transmitted or reflected light at 64X-100X magnification. Generally the albumen gland tissue was teased away, the ducts of the uterus and prostate separated, then the isolated carrefour area placed in a temporary glycerine mount on a depression slide for study. The species for which such detailed observations were made are Opanara duplicidentata (fig. 103a, b), Taipidon petricola petricola (fig. 138b), Endodonta fricki (fig. 164b), Libera micrasoma (fig. 171g), Thaumatodon euaensis (fig. 195c), and Aaadonta constricta constricta (fig. 199c). Inspection of these figures shows that the talon inserts either directly on the head or slightly to one side of the head of the carrefour. The hermaphroditic duct uniformly inserts laterally on the side of the carrefour, either just below (O. duplicidentata} or well below the carrefour head. The albumen gland duct is drawn as entering opposite the hermaphroditic duct in Opanara duplicidentata, but this needs confirmation. Since it was the first of these species studied and illustrated, this structure could have been misinterpreted. The other species show the albumen gland duct entering after the uterine duct has separated from the prostate. In Libera micrasoma and Aaadonta c. constricta it was possible to indicate the internal duct passageways. These appear in transparency to be simple tubes. The situation in the studied Charopidae is quite comparable, with different species having short or long talon stalks, but generally the talons of charopids have a distinctly circular head, resembling a golf ball on a tee, while that of the Endodontidae is elongately ovate, and there is a greater tendency for the carrefour not to show any expansion in the Charopidae. Histological studies are needed to work out the significance of these reported differences and to establish the exact structures involved. ALBUMEN GLAND (GG) - This mass of alveolar tissue is situated just above the pallial cavity apex and nestled in the loop of the intestines just below or at the point where the narrow esophagus widens to form the stomach. The albumen gland varies greatly in size from individual to individual, often has its surface indented by the head of the spermatheca and intestinal loops, and fragments upon handling. Most typically, whether the snail is contracted or expanded, a well-developed albumen gland lies next to the upper parietal wall. It will be broadest at the base, tapering to a narrow rounded tip just before the stomach expansion reaches the upper parietal margin. The spermathecal head lies at its base, next to the parietal wall, and between the albumen gland base and the kidney base. The intestinal loop that indents the kidney margin lies between the spermathecal head and the palatal wall. When reduced in size, the albumen gland may lie in the midwhorl section rather than being pressed against the parietal wall. Because of its variability in size, the necessity to remove tissue in order to examine the talon, and its general delicacy of structure, no attempt was made to show the albumen gland in complete form. Compared with the Charopidae, the alveoli were smaller and the shape less variable in the Endodon- tidae. Charopid taxa have the albumen gland tightly jammed between the intestinal loops with the surface deeply indented by other structures. The gland has distinctly larger alveoli and is more ovoid than elongated in shape. Probably this relates to the fact that the Charopidae from Pacific Islands have a median whorl count of 4!/s- compared with a median of 5!/2 + for the Endodontidae. Shortening of the visceral hump coiling resulted in less space for "stringing out" the organs and hence greater compaction in the area between the pallial cavity apex and the stomach expansion. PROSTATE (DG) and UTERUS (UT) - Most European workers have preferred the term "spermovi- duct" for this region, reflecting the fact that in most European pulmonates there is a morphologically united common duct that is physiologically differ- entiated into male and female tracts. Good descrip- tions of the tract histology for this area in advanced land snails are given by Van Mol (1972, pp. 202-207) for the Bulimulidae and by Rigby (1965) for the Succineidae. The Charopidae agree with this pattern in having the ducts fused, but with a simpler pattern of differentiation than was seen in the above mentioned groups (Solem, unpublished). The Endodontidae differ in having the prostate and uterus entirely separate tubes that are very loosely bound together by a few connective tissue strands. Elsewhere (Solem, 1972b, pp. 108-112) I have reviewed the pattern among pulmo- nate snails of variation in fused versus separate pallial gonoducts ( = prostate and uterus), and concluded that fully separated ducts are the basic pulmonate condition, with fusion of the tracts a derived character state and advanced condition that has evolved several times. Throughout the Endodontidae, the prostate is a small, thin-walled tube into which one (fig. 191a), two (fig. 199c), three (fig. 164d), or even four or five (fig. 165a) rows of alveoli attach individually. Normally the prostate lies on top of the uterus, with the sperma- thecal shaft lying loosely attached on top of its duct. The prostatic alveoli are bent toward the outside of the whorl across the top of the uterus. The whole bundle of genital ducts (prostate, uterus, spermathecal shaft), esophagus, and columellar muscle occupy the lower palatal and columellar portion of the whorl margin, while the pallial cavity extends to the palatal margin. The hindgut occupies the upper parietal- palatal margin. In the Charopidae, the prostate tissue is fused to the uterus, with an internal channel slightly to strongly demarcated from the uterine lumen. The prostatic tissue is composed of even fewer and often 78 SOLEM: ENDODONTOID LAND SNAILS more elongated alveoli that show no patterned arrangements, rather than the large, rather short alveoli placed in even rows that are characteristic of the Endodontidae. Structure of the uterus in the Endodontidae is far more complicated, but the histological studies needed to work out the details could not be undertaken. Depending upon the species, there is evidence of from two to four sections in the uterus. These range from an always present division into a simple, thin-walled, narrow upper chamber (fig. 164a, UTi) and a broader, thicker walled, glandular chamber (UT:>) in the lower section, to the situation seen in Thaumatodon hys- tricelloides (fig. 191b). This species has a clump of glandular tissue, clearly different in texture from that of the albumen gland, around the head of the uterus (UTi); the typical thin-walled, rather narrow tube (UT-j); the expanded zone with thick glandular walls (UT;i); and a narrower tube with very fine glandular tissue (UTi) that lies below the termination of prostatic alveoli and before the start of a simple tube that I interpret as the free oviduct (UV). Presumably the upper zone of the uterus is the section associated with enclosure of the embryo with nutritive material and membranes, while the lower glandular section secretes the egg capsule. No eggs were seen inside the uterus of any specimen dissected during this study. Whether the more complex uterine area of T. hystricelloides reflects a change in structure or merely better preservation (this material initially was placed in 95 per cent alcohol after drowning) is unknown. The uterine area tended to swell quickly in liquids of lowered alcohol content and few detailed observations were made on this area, since after initial illustration and handling it often had burst or split open. In the Charopidae, the uterus is much more sharply divided into two zones, with the lower chamber having thick glandular walls and often appearing as an ovoid structure grafted onto the prostate surface. The difference between having totally separate versus fused pallial gonoducts is a highly significant one and another key character separating the Endodontidae from the Charopidae. Inspection of the drawings suggests that there may be differences in the number and actual size of the prostatic alveoli, but whether this is a variation correlated with reproductive cycle stage or has phyletic significance is unknown. TERMINAL MALE GENITALIA - This complex includes the vas deferens (VD), epiphallus (E) (when present), penis (P), penial retractor muscle (PR), and the atrium (Y). Differences are substantial, primarily because species recognition presumably has great selective influence on this region. Because of this, description of the structural variations cannot be separated completely from a consideration of sym- patric interspecific variation. The vas deferens, when prostate and uterus are separate, must be defined as the prostatic duct below the end of the alveoli. Its appearance varies with preservation and possibly between taxa, but data are meager on this latter aspect. In some taxa it appears that the tube becomes thicker walled and reflects light more strongly almost immediately, but in others this change seems to occur gradually as the vas deferens descends to the penioviducal angle. The vas may or may not be loosely bound to the angle before reflexing upward along the penis itself. It is effectively held in the angle by the right ommatophoral retractor muscle that passes through the penioviducal angle. The ascent of the vas deferens is on the columellar side of the penis to a point that is usually distinctly below the head of the penis (fig. 46), although the exact point of entry varies within rather narrow limits for any single lineage. The entry into the chamber of the penis is a simple pore, normally distinctly below the point of apical union for the pilaster. The situation in Endodonta (figs. 164a, c; 165a, d, g) is typical. Taipidon p. decora (fig. 49a) shows an exceptionally low entrance. The basic two pilasters of the penis chamber unite at its apex and while drawing tech- FIG. 46. Vas deferens entrance and penial retractor muscle insertion patterns: a, typical (based on Cookeconcha jugosus); b, apical vas deferens entrance (based on Ruatara oparica normalis)', c, some added tissue to penis apex (based on Opanara megomphala); d, moderate extra tissue (based on Opanara bitridentata); e, much extra tissue (based on Planudonta intermedia); f, formation of a penial epiphallus and valvular vas deferens entrance (based on Aaadonta c. contricta). (CK) PATTERNS OF MORPHOLOGICAL VARIATION 79 niques may suggest that the pore (EP) lies opposite the point of union or on the same side as the point of union in generalized taxa, this is an artifact of drawing technique rather than an indication of structure. In forms with altered pilaster arrangement, such as Cookeconcha hystricellus (fig. 165j, k), the pore often is located just below the point of pilaster union, although externally (fig. 165j) it may appear to be apical in its insertion. In Taipidon fragila (fig. 138e) the penis is only 1.5 mm. long, compared with a 2.5 mm. long penis in its sympatric congener, T. vari- dentata. The former has a distinctly more apical insertion of the vas than does the latter. In Rhyso- concha variumbilicata (fig. 64f, g) the vas deferens entrance is almost apical, and in R. atanuiensis it is apical. The two dissected races of Ruatara oparica (figs. 46b; 64h-j) have the vas entering "just below the apex." These situations probably are indicative of size reduction. The above changes represent positional minor shifts. In contrast, there is a major change in the Thaumatodon-Priceconcha-Aaadonta lineage. In these genera the vas deferens enters through a pair of "lips" or a "valve" (fig. 46f; 199d) into a reflexed zone of the penis that lies well below the point of attachment for the penial retractor muscle. Two pilasters extend from this point, in varying degrees of prominence (figs. 199d; 200c; and Solem, 1973d, fig. 20, b), up to the penis head and then into the main chamber. The length of the reflexed portion varies, being about one-third the penis length in Thaumatodon decemplicata, T. hys- tricelloides, and T. spirrhymatum, one-half the penis length in T. euaensis and all dissected Aaadonta, and two-thirds the penis length in Priceconcha tuvu- thaensis (Solem, 1973d, fig. 20, a). The homologies of this area and appropriate terminology present some difficulties. An epiphallus usually is defined as an expanded portion of the vas deferens that is functionally apical of the penis head. Normally, in advanced taxa, it is involved in sper- matophore or sperm packet formation. Presumably in less advanced taxa it secretes fluids to carry and provide life support for the sperm during transport. It is not eversible and is very different from the penial structures. Frequently, in such groups as the Helica- rionidae (Solem 1966a, figs. 6, b; 19, a; 22, a), there are accessory flagella or caeca on the epiphallus, and in Indian Helicarionidae (Blanford and Godwin-Austen, 1908) highly complex and spined spermatophores are formed in the epiphallus and its appendages. It also is possible for an epiphallus to develop from the penis itself and thus be a non-eversible apical chamber of the penis. The latter situation probably applies to the Thaumatodon- Aaadonta lineage. The altered zone lies topographically below, although functionally above, the penial retractor muscle and is reflexed anteriorly in all specimens. It is extremely difficult to see how this could be everted successfully. If this area is not everted during copulation, it would be functionally a "penial epiphallus." On the other hand, the reflected, probably non-eversible area has the typical penial pilasters found in the other Endodontidae. It is clearly of penial origin and probably of recent origin. The valvular arrangement at the point of vas deferens entrance is the "new" structural feature, while the change in penial retractor attachment, or equally possibly the growth of the penial chamber past the retractor muscle attachment, is a minor shift in position. The pilaster patterns are typical of the family. I have concluded that this is a functional penial epiphallus of recent origin and use the term epiphallus for the structure throughout this report and in an earlier publication (Solem, 1973d). The adoption of a valvular entrance from the vas deferens and alteration of the upper penial chamber into an epiphallus represents the most striking change seen in the genital anatomy of the Endodontidae. In contrast to this, the Charopidae normally have a vas deferens-derived epiphallus, with the entrance from the vas highly complex and variable, often involving sphincter pilasters and/or plug-like struc- tures. The epiphallus normally is sharply differ- entiated from the penis and there are highly complex and varied patterns of transition between the two areas. In lineages that are undergoing visceral hump reduction in the Charopidae, such as the "flammulinid" genera Flammulina and Maoriconcha, the epiphallus may be shifted forward into the penial complex, thus giving an external appearance of no epiphallic differentiation (Solem, unpublished). One group of Micronesian Charopidae departs from the pattern in having direct entrance of the undifferen- tiated vas deferens into the penis head after passing partway through the U-shaped insertion of the penial retractor muscle (Solem, unpublished). There is no trace of any epiphallic differentiation. The penis in the Endodontidae is basically an elongated tube with two pilasters extending from a union at the apex of the penis almost to or even into the atrium. Differences in penis length, pilaster pattern, and structure of the pilasters are highly influenced by interactions with sympatric congeners. Discussion of these factors follows preliminary review of some phyletic alterations. The Marquesan genera Taipidon and Planudonta are unique in the family because of developing a pustulose zone in the penis (figs. 49b; 138h; 147b, d). The extent and position of this glandular zone differs in the various species, ranging from perhaps two-thirds of the penis length in T. semimarsupialis to just the basal area in T. petricola decora. This is an additive feature to the genitalia and has no equivalent in other geographic areas. Attachment of the penial retractor muscle (PR) to the penis itself is variable (fig. 46). In all dissected Minidonta, Mautodontha, Cookeconcha, Rhysoconcha, 80 SOLEM: ENDODONTOID LAND SNAILS Ruatara, Australdonta, Nesodiscus, Nesophila, Endo- donta, and Libera the penial retractor attaches directly onto the penis apex (fig. 46a). Upon opening the penis there is no indication of any extra thickness to the penis apex at or near the point of muscle attachment. In Orangia, Planudonta, several Taipi- don (T. centadentata, T. fragila, and T. varidentata), and all Opanara except O. depasoapicata and O. caliculata, there is a "fleshy extension" to the penis apex that extends for a variable distance upward before the penial retractor attaches (figs. 46c-e). This partly glandular zone does not surround the muscle, but is a new zone between the muscle and the penis apex. The varying extent and prominence of this zone within a genus can be seen quite clearly in Figures 96 and 97. The typical apical insertion of O. depasoapi- cata (fig. 96a, b), weak extension in O. altiapica and O. megomphala (fig. 97a-d), moderate extension in O. m. tepiahuensis, O. fosbergi, and O. perahuensis (fig. 97e-h), and stout extension in O. bitridentata (fig. 96e, f) compare with a much more even extension in the species of Orangia (fig. 121). Taipidon shows great variability, with direct attachment in T. petricola (fig. 49a) and T, semimarsupialis (fig. 139a, b), a long but narrow attachment zone in T. centadentata (fig. 139e, f), and very prominent attachment zone in T. fragila, T. varidentata (fig. 138e, g), and the Planudonta (fig. 147). In all of the above taxa, both the union of the pilasters and apex of the penis lie a significant distance below the penial retractor attachment point. The function of the added zone is not known, but a logical possibility for investigation would be secretion of fluids equivalent to that normally provided by the epiphallus in more advanced taxa. The different patterns of penial retractor attach- ment and vas deferens entrance are summarized in Figure 46. In essence, these seem to represent at least two separate kinds of experiments in providing the functional equivalent of an epiphallic zone: inter- position of glandular tissue between the penis apex and penial retractor muscle in several groups, and then formation of a non-eversible section of the penis by reflexion in the Thaumatodon-Priceconcha-Zyzzyx- donta-Aaadonta complex. The addition of glandular tissue to the penis apex is found in Rapan and Marquesan genera, but not in taxa from other geographic areas. Much more variability is seen in the shape of the pilasters. This is complicated in analysis by changes in relative size shown by the two basic pilasters, and some variability in pilaster number. Figure 47 shows the basic cross-sectional pilaster types: low and rounded (a); medium in elevation (6); or high and lamellate (c). In Australdonta (fig. 125b, e) the two pilasters are complexly folded and split into portions, so that they do not fit into any of the above categories. In Ruatara (fig. 64i) there is only a single rugose pilaster for most of the penis length, which becomes bifurcated basally, suggesting that it is a secondary modification from the basic two-pilaster pattern. The low and rounded pilasters are characteristic of the few Minidonta, Mautodontha, Endodonta, and Aaadonta that have been dissected, plus such species as Cookeconcha jugosus (fig. 165h), Taipidon centa- dentata (fig. 139f), and Planudonta intermedia (fig. 147b). Those taxa in which the pilasters are higher than wide, but rounded above if not secondarily altered, include all Thaumatodon, Planudonta sub- planula (fig. 147g), Nesophila tiara (fig. 165g), Neso- discus (fig. 154), the other Cookeconcha, Rhysoconcha, Opanara perahuensis (fig. 97i), and Orangia sporad- ica (fig. 121k). The latter two species probably represent secondary modifications from the normal pattern in these genera, since the remaining Orangia and Opanara all have high, lamellate pilasters. These also are the normal condition in Libera, the remaining Taipidon, Planudonta concava, Kondoconcha othnius (p. 368), and Priceconcha tuvuthaensis (Solem, 1973d, fig. 20, b). Whether the pilasters unite both apically and before the atrium, as in Endodonta; extend routinely into the atrium, as in Thaumatodon and Aaadonta; or are variously altered is selected for at a low taxonomic level. There is a general tendency for those taxa with low pilasters to have union apically and basally, while for those with very high and lamellate pilasters there can be absence of any union (Libera, fig. 171b) or typical union apically (Plan- udonta, fig. 147b, d; and Priceconcha, Solem, 1973d, fig. 20, b). FIG. 47. Pilaster cross-sectional patterns in typical Endodon- tidae: a, low and rounded; b, elevated but rounded; c, high and lamellate (main pilaster) with reduced pilaster flattened or otherwise altered. (CK) PATTERNS OF MORPHOLOGICAL VARIATION 81 The relative size of the two pilasters is partly phyletically correlated and partly influenced by sym- patric factors. The two pilasters are equal or nearly equal in size for dissected Minidonta, Mautodontha, Rhysoconcha, Australdonta, Nesodiscus, Nesophila, Endodonta, and all Libera except L. cookeana. Species of Orangia, Planudonta, Thaumatodon, Price- concha, Aaadonta, and Taipidon (except two special cases) have the pilasters grossly unequal. Species of Cookeconcha and Opanara have a nearly equal mixture of both types. Ruatara oparica (fig. 64i) is unique in that the pilasters apparently fused into a single rugose structure that is split basally. There is no significant correlation between pilaster shape and relative size of the two pilasters (table LI), although a tendency exists for the low and rounded pilasters to be equal in size, with the higher pilasters more likely to be unequal. TABLE LI. - CORRELATION OF PENIAL PILASTER SHAPE AMD RELATIVE SIZE Pilaster shape Low and rounded Intermediate High and lamellate Pilaster relative size Equal Subequal 9 6 15 The gross data on this variation should be checked with the patterns of sympatric variations summarized in Tables LII through LIV. Unfortunately, I had available for study only one example of sympatric congeners where low and rounded pilasters were involved. These two Tahitian Mautodontha (table LIII) differed grossly in penial size, but not in pilaster pattern. Their shell features (fig. 74) also are grossly different, so that there is no question of their being valid species. With high and lamellate pilasters, obviously there is much more room for ex- perimentation. Overall the genus Opanara has six taxa with unequal pilasters and five with essentially equal pilasters. Listing several sympatric Opanara (table LII) shows that both penis size and pilaster changes occur. Orangia shows equivalent shifts. In the Tahi- tian Libera (table LIII) there is still a relatively simple pattern of penis size and pilaster proportion shifts, but in the Marquesan species (table LIV) the addition of a pustulose zone in the penis increased the options for change. The two Planudonta show size, pilaster length, and pustulose zone location variations, while the Taipidon show a unique pattern of change. On two different islands sympatric congeners have diverged in penial structure by one species having the pilasters broken up into a series of low bumps (figs. 138f, h; 139b, f). The relatively few examples cited in these tables represent all the situations in which the anatomy of sympatric congeners could be studied. They have a greater importance by demonstrating the extent and ways in which these structures can be altered by interspecific selection phenomena. They also suggest that the size and shape variations seen in the penes of the few Endodonta and Cookeconcha dissected during this study (figs. 164, 165), none of which are sympatric, do not have great systematic importance. Rather it is probable that these variations hint at the extent of sympatric diversity in the essentially unstudied Ha- waiian endodontid fauna. Similarly, they suggest that the wild variations seen in the terminal genitalia of New Zealand and New Caledonian taxa have to be interpreted in terms of such character displacement reactions. At the same time, a note of caution must be injected here. Some of the outlined situations may be oversimplified. In the species accounts of Libera bursatella bursatella and L. b. orofensis reference is made to differences between and within populations in regard to penis size. Whether this is dimorphism, sibling species, or an apparent reversal of character displacement under conditions of exact sympatry is unknown. The L. b. bursatella with generally longer penes came from Station 863, where L. cookeana (penis length 5.9 mm.) occurred, but not L. micra- soma, with a penis length of 3.9-4.1 mm., which occurred with L. b. bursatella (penis length 4.3-4.5 mm.) at Station 866. Since there are clear-cut pilaster differences between the three species, the possible tendency toward closer similarity in penial length with sympatry of the different species is of very uncertain significance. Further work on these problems will depend upon new collecting efforts, since yet another complication exists. At Station 863, there were 169 L. b. bursatella and only one L. cookeana, while at Station 866, there were 90 L. b. bursatella and six L. micrasoma. The extent to which such disparate numbers would affect character displacement is unknown. The Pacific Island Charopidae normally have a verge or vergic papilla and sophisticated circular or pocket-like stimulatory pads or pilasters. In the absence of these structures (epiphallate Micronesian genera), there is a combination of added muscular sheaths to the penis and epiphallus, plus a radiation of numerous fine lamellar pilasters from the epiphallic pore into the penis interior, where they coalesce into three large longitudinal pilasters. Origin of the penial retractor muscle is yet another variable feature in the Endodontidae. In 29 of the 40 taxa in which this could be observed, the muscle originates from the diaphragm. While in one case the penial retractor is attached as low as the middle of the pallial cavity (Aaadonta fuscozonata) normally it attaches opposite the pallial cavity apex. A. fuscozonata has a sharply increased whorl count in comparison with other Aaadonta, so there may have been a forward shift of the muscle. In Aaadonta kinlochi and A. constricta, the attachment is just below the apex. It is slightly above the apex in Orangia cookei and Opanara areaensis, while in 82 SOLEM: ENDODONTOID LAND SNAILS Minidonta hendersoni the penial retractor attaches just at the point of stomach expansion. This variability serves to indicate how gradual apical shift of the penial retractor origin could lead to the transfer of the muscle from the diaphragm to the free muscle system, generally the columellar muscle. This state is known in 11 taxa, including all dissected Endodonta and Australdonta. Both Planudonta for which the origin of the retractor muscle is known have it coming from the columellar muscle. The same state is found in both subspecies of Taipidon petricola, the smallest species in its genus, although other Taipidon have the more normal retractor muscle origin. In Opanara duplicidentata the penial retractor attaches to the tail fan just before this is joined by the buccal retractors. In Nesodiscus fictus the penial retractor TABLE LIT. - SYMPATRIC VARIABILITY IN PENIS SIZE AND PILASTER PATTERNS IN RAPAN TAXA Species Penis length in nun . Opanara bitridentata 2.65 duplicidentata 1.5 areaensis 2.65 megomphala 2.0-2.7 altiapica 1.8-2.1 Orangia cookei 2.0-2.1 sporadic a 2.3 maituatensis 3.6 Ruatara oparica 3.3 Rhysoconcha 1.6-1.9 Pilaster pattern both equal one split one reduced both equal one greatly reduced unequal, both high, one split subequal, both widened, shorter subequal, both split only one corrugated, split basally both equal attaches partly to the columellar muscle and partly to the diaphragm. This shift of the penial retractor muscle origin has no obvious correlatives, but does have a significant result. Since the columellar muscle extends much further apically than the diaphragm, which attenuates as connective tissue at the level of the stomach, this shift in origin permits lengthening of the penis (fig. 56). A long penis is not a prelude to this shift, since Opanara duplicidentata, the only species in that genus to show the change in penial retractor origin, also has the smallest penis in the genus. In both Planudonta and two of the three Endodonta, penis length exceeds the shell diameter. In most other taxa it is considerably less. Only Taipidon semimarsupialis of the taxa with diaphragm origin of the penial retractor has the penis longer than the shell diameter. This probably is an accident caused by the secondarily changed growth pattern (fig. 143e) associated with brood-chamber formation in that species. In Libera micrasoma shell diameter and penis length are TABLE LIII. - SYMPATRIC VARIABILITY IN PENIS SIZE AND PILASTER PATTERNS IN TAHITIAN SPECIES Species Mautodontha zimmermani aoraiensis Penis length in mm. Libera micrasoma b_. bursatella cookeana 3.9- 1 *.! U.3-U.5 5.9 Pilaster pattern both equal both equal equal , simple subequal , folded one greatly reduced and only on lower half of penis TABLE LIV. - SYMPATRIC VARIABILITY IN PENIS SIZE AND PILASTER PATTERNS IN MARQUESAN SPECIES Species Planudonta intermedia concava Taipidon centadentata semimarsupialis fragila varidentata Island N N N N H H Penis length in mm. U.6 6.0 U.3 *. 9-5. 3 1.5 2.5 Pilaster pattern small pilaster shorter, pustulose zone lower in penis small pilaster longer , pustulose zone more medial split into bumps unequal, high unequal, high split into bumps N=Nukuhiva ; H=Hivaoa PATTERNS OF MORPHOLOGICAL VARIATION 83 virtually identical. This is another brood chamber situation. In the Pacific Island Charopidae, the penial retractor muscle inserts in a U-shaped fan on the penis-epiphallus junction, on the head of the highly complex epiphallus in the epiphallate Micronesian genera, or on the head of the penis with the vas deferens passing through the muscle in the third major group of taxa. The atrium (Y) is formed by the union of the penis and terminal female genitalia (spermatheca, vagina, free oviduct) to form a common channel to the gonopore. Generally this is quite short, but in Australdonta, Endodonta, and Aaadonta the atrium is noticeably longer than normal, and in Minidonta hendersoni, Taipidon varidentata, Planudonta con- cava, and Nesophila tiara there is some lengthening present compared with related taxa. The above discussion of variations in the male terminal genitalia can be summarized rather simply. The vas deferens normally enters the penis laterally through a simple pore, but in a few taxa more apical entry is obtained. In one lineage, the Thaumatodon- Aaadonta sequence, experimentation in development of a penial epiphallus has been accompanied by adoption of a valvular vas deferens entrance. In other groups accessory tissue on the penis apex may serve an epiphallic secretory function. The penis itself normally has two longitudinal pilasters that may be low and rounded, elevated and rounded, or very high and lamellate. The pilasters may be equal in size or sharply different, frequently depending on character displacement interactions with sympatric congeners. Major changes in Ruatara and two species of Taipidon apparently result from this phenomenon. The penial retractor muscle normally attaches to the diaphragm, but can shift to the columellar muscle just above the point where it is formed by union of the tail fan and buccal retractors. A few genera show a significantly elongated atrium, but most are short. The only additive structures to the terminal male genitalia are the pustulose zone found in the two Marquesan genera: the valvular entrance of the vas deferens seen in Thaumatodon, Aaadonta, and Priceconcha', and the added tissue on the penis apex in Orangia, Planudonta, many Taipidon, and some Opanara. The other variations are quite minor modifications on a unitary theme. The Charopidae present a considerable contrast in having an epiphallus originating from the vas deferens, frequently a verge or vergic papilla, usually quite complex and extensive pilaster arrangements in the penis, often a very short penial retractor muscle that inserts in quite a different fashion, and radical changes in structure from group to group of the Pacific Island taxa. Even the isolated penis would be sufficient to indicate if a species belonged to the Endodontidae or Charopidae, and often what genus or group of genera. TERMINAL FEMALE GENITALIA - This includes the post-uterine tubes, which are quite simple and uncomplicated in the Endodontidae. The normal pattern in land snails is for a thin-walled or muscular tube to lead from the glandular uterine section of the spermoviduct (if it shows a fused prostate-uterus) or uterus for a short-to-medium distance. This is the free oviduct (UV) and may be highly complex internally. It is joined by the shaft of the spermatheca (S). The combined tubes then extend forward to the atrium as the vagina (V). In the Charopidae, for example, the vagina may have almost as complicated a set of pilasters as is found in the penis of that group, the lower part of the spermathecal shaft may be enor- mously swollen because of high pilasters extending from the upper vagina, with the free oviduct opening reduced to a narrow pore. The upper portion of the free oviduct may have thick walls and be glandular in nature, while the spermathecal shaft narrows to a thin tube at the anterior end of the uterus, ascending along the latter to a typically ovate expanded head. There are numerous characters to use in systematic analysis, and in taxa from New Zealand and Australia it seems probable that the terminal female genitalia, as well as the male genitalia, are involved in species recognition. The Endodontidae, in contrast, have extremely simple terminal female genitalia. A thin tube, without prominent internal features, extends from the lower section of the uterus to the atrium. It may be joined by the spermathecal shaft (which is a uniformly undifferentiated tube for its entire length) slightly to moderately above the atrium, thus technically forming a short vagina, or the spermatheca may insert on the penioviducal angle, effectively entering the atrium directly. In a few taxa, mainly the Thaumatodon- Aaadonta assemblage, the spermatheca actually in- serts on the penis itself, slightly above the atrium. Regardless of where the spermatheca inserts, there are no indications of structural differentiation in the female tubes, nor even recognizable pilasters. They are extremely simple ducts. The pattern of spermathecal insertion is primarily phyletic, with some convergences having occurred. Penial insertion of the spermatheca (fig. 48a) is characteristic of Thaumatodon, Aaadonta, and Price- concha. The entrance is on the inside of the penioviducal angle, distinctly above the point where the penis separates from the atrium. There is no special internal structure evident. In Rhysoconcha (fig. 64g) the same pattern is seen. Atrial (fig. 48b) insertion of the spermatheca is normal in Minidonta, Austral- donta, Orangia, Libera, Taipidon, Planudonta, all Opanara except two, and Mautodontha aoraiensis. Oviducal insertion (fig. 48c) occurs in Nesodiscus, Nesophila, Cookeconcha, Endodonta, Ruatara, Opan- ara depasoapicata, O. bitridentata, and Mau- todontha zimmermani. The distance shift is not great and I question whether this makes any functional significance. Atrial and oviducal insertion occurs in 84 SOLEM: ENDODONTOID LAND SNAILS FIG. 48. Spermathecal insertion patterns: a. penial insertion (based on Thaumatodon euaennin); b, atrial insertion (based on Libera coakeana); c, oviducal insertion (based on Endodonta fricki). (CK) both Mautodontha and Opanara, but otherwise there is phyletic correlation. In all Endodontidae and Charopidae the sperma- thecal head extends past the pallial cavity apex. It is an elongately bulbous structure that lies next to the base of the albumen gland (fig. 49a) along the parietal wall and then next to the kidney base. It is sometimes lost during dissection of charopids because the artery passes over the spermathecal stalk to bind it into the apical viscera, but in all Endodontidae this appears to be free of this loop and thus dissects out easily. In order to show the relationships of this organ, Endo- donta fricki was drawn with the organs pulled slightly apart. The perspective in this drawing (fig. 163d) might be interpreted as indicating that the spermathecal head lies below the pallial cavity apex, but this is an artifact. Many spermathecae seen in this study contained compact masses of sperm, but these were not enclosed in a membrane. There is no evidence of any spermatophore formation. The above discussion of the genital system surveys patterns of variation and features observable after teasing apart the separate organs. It is useful to have one drawing in which the organs are portrayed as they are dissected out in early stages of study. Figure 49a attempts to show the just-excised genitalia. The material of Taipidon petricola decora was received for study long after the main systematic section had been completed. Hence it was decided to draw this showing the typical position of the organs when first dissected out, rather than as separated, to show all origins and insertions. This permits showing one apparently quite characteristic feature of the family. The shaft of the spermatheca (S) lies first on top of the free oviduct, then it sits directly on top of the vas deferens (VD) and prostatic duct for about the lower half of the prostate (DG). About the middle of the prostate, the spermathecal shaft shifts in position to lie on top of the prostatic acini and parallels the prostatic duct for the upper third. This is the typical position for the spermathecal shaft. The penis (P) has been shifted laterally in this view and is unusual primarily for the very low entrance of the vas deferens. The above account concludes discussion of the structural and topographic variations seen in the genital anatomy of the Endodontidae. The differences from the Charopidae are substantial and obvious if even a fragmentary part of the genitalia is available. These are summarized in Tables LVIII, LIX following discussion of the other organ systems. Discussions of some size correlated variations in anatomical proportions are deferred, since several organ systems are involved. Pallial complex In the Endodontidae this area shows a simple and relatively uniform structural pattern, whose variations are mostly correlated with size, whorl count, and features of the palatal barriers. The typical configuration is shown in Figure 49c. A more detailed view is presented in Figure 195a. The most significant features for phylogeay are the weakly bilobed kidney (K) that reaches the hindgut (HG), the short reflexed ureter (KD) that opens in a ureteric pore (KX) just at the anterior tip of the rectal kidney arm, the simple pulmonary vein (HV), and the variable pattern of a mantle gland extension onto the pallial roof being present (fig. 171e) or absent (fig. 49c). The total PATTERNS OF MORPHOLOGICAL VARIATION 85 absence of any secondary ureter or urinary groove leading from the ureteric pore (KX) to the pneumostome is a feature of major phyletic impor- tance (p. 103). The functional significance of this is that ex- cretory products must exit onto the pallial roof surface. Water must then be used to flush out the waste products. In the typical sigmurethrans there is either a highly vascularized channel (many Bulimu- lidae and other holopodopid taxa) or a closed secon- dary ureter (Limacacea, Camaenacea, Helicacea) through which the excretory products pass. This divergence in the Sigmurethra was pointed out by H. B. Baker (1962a; 1963, p. 220). The secondary ureter seems to function primarily in water resorption. Elsewhere I (Solem, 1974) have postulated that the evolution of a secondary ureter was a necessary preadaptation to the evolution of land slugs from shelled ancestors. The exact level of classification required to recognize the third type of structure seen in the Endodontidae, a closed primary ureter, but no trace of any secondary groove or tube, cannot be considered here. It is quite probable that the difference in median mean whorl count between the Endodontidae (5'/2+) and Charopidae (4'/8 ) relates to the fact that the former lacks a ureter, and thus needs a greater pallial cavity area for holding a water reservoir (Blinn, 1964) than does the Charopidae in which there is a secondary ureter present. The only charopid known to me that may lack a secondary ureter is the Tasmanian species Planilaoma luckmanii (Brazier, 1877). The material available of this species was very limited and this observation (Solem, unpublished) needs to be confirmed by more dissections. To date, this is the only Indo- Pacific taxon seen that might be in any way transitional from the pallial cavity states of the Endodontidae to the Charopidae. The normal pattern in the Charopidae is for an evenly bilobed kidney (secondarily altered with accen- tuation of either rectal or pericardial arms in different taxa); a complete secondary ureter opening lateral to the anus at the pneumostome; only very rarely (in Micronesian and Melanesian, but not New Zealand taxa) any mantle gland invasion of the pallial roof; the rectal kidney arm extensively overlapping the hindgut (often both dorsally and ventrally); and the heart much more deeply indenting the kidney. The differ- ence of the secondary ureter in the Charopidae is the fundamental distinction. This means that even the smallest anterior fragment of the pallial cavity is sufficient to decide to which family the species belongs on the basis of ureter presence (Charopidae) or absence (Endodontidae) by the hindgut just inside the pneumostome. Variation in the pallial region concerns the presence or absence of a glandular extension from the mantle collar onto the pallial roof, the relative length FIG. 49. Anatomy of Taipidon petricola decora. North side of Vaituha Valley, 600 ft. elevation, Eiao, Marquesas. BMNH: a, genitalia; b, interior of penis; c, pallial structures. Scales lines equal 1 mm. (CK). (See Appendix for explanation of abbreviations.) and prominence of the kidney arms, and the actual length of the cavity itself. The correlatives of these changes, which recur sporadically throughout the family, are not strong. The mantle collar extension involves shifting of some glandular materials from the collar itself posteriorly onto the pallial roof. The thickness and obvious nature of the extension varies rather widely, perhaps indicative of the stage in glandular activity. This area presumably is involved in the secretion and resorption of the palatal barriers. In swollen state this could be indicative of recent posterior barrier resorption, while in reduced state this could be indicative of recent deposition of calcareous material. Thus variation from individual to individual could be expected. Because specimens had been preserved mostly for long periods in alcohol of uncertain strength and acidity, retention of calcium residue in the glandular areas could not be depended on and no tests for calcium were made. The mantle 86 SOLEM: ENDODONTOID LAND SNAILS collar extensions were obvious in Australdonta de- gagei, but not in A. raivavaeana; prominent in Endodonta (fig. 163e) and Nesodiscus (fig. 154b); prominent (fig. 171e) to absent (fig. 172e) in the illustrated Libera (but clearly present in other examples of Libera fratercula); present in most Thaumatodon and all Aaadonta', and also seen in Taipidon semimarsupialis. No traces of any pallial roof glands were seen in specimens of Minidonta, Cookeconcha, Opanara, Rhysoconcha, Orangia, the other Taipidon, Planudonta, or Nesophila. Many of those taxa with mantle roof gland extensions have long and/or deeply recessed barriers (10 taxa). The shift in mantle gland tissue under these conditions has obvious value in permitting barrier resorption and deposition activities with less complete animal retrac- tion. Taipidon semimarsupialis and Thaumatodon euaensis, in contrast, have short (one-eighth whorl) barriers at the lip edge, the two Australdonta and Endodonta lamellosa have equally short, slightly recessed barriers, while the other two Endodonta have moderately recessed, short barriers. Taxa with such long barriers as Priceconcha (Solem, 1973d, p. 23) show no trace of mantle gland extension. While 10 taxa with mantle gland extensions have extra long and/or deeply recessed barriers, six do not, and some taxa with deeply recessed or very long barriers lack any trace of such an extension. The variation in kidney width was not quantified, but obviously varies, as can be seen by comparing the very narrow, folded kidney of Libera b. bursatella (fig. 171d) with the far broader kidney of Aaadonta c. constricta (fig. 199a) and the average kidney shape of Thaumatodon euaensis (fig. 195a) or Taipidon petri- cola decora (fig. 49c). The broadened kidney in Aaadonta probably correlates with the comparatively great palatal wall distance from the parietal-palatal margin to the high second palatal barrier (fig. 203b), while in T. euaensis (fig. 194b) and T. petricola decora (fig. 140c) this same zone is distinctly narrower. Barriers would not be a factor in Libera b. bursatella, but the extreme deflection of the last few whorls (fig. 174a) has shortened the upper palatal wall (that portion above the weak peripheral keel) quite dras- tically. The extreme narrowing of the kidney in this species can be a byproduct of this space change. Relative length of the kidney lobes was scored as the rectal lobe (that next to the hindgut) being distinctly less than half the length of the pericardial lobe (that next to the heart), or essentially half the length. Often the difference is quite striking, as in Aaadonta constricta and A. fuscozonata (figs. 199a, f), even though both of these are scored as having the "short" rectal lobe. The longer type rectal lobe is found in a variety of taxa, such as Nesodiscus (fig. 154b) and various Libera (figs. 171d; 172e) with elongated pallial cavities and high whorl counts, but also in Opanara, Ruatara, and some Cookeconcha with typical or possibly slightly reduced whorl counts. HG FIG. 50. Pallial cavity length variation, based on the typical (a) and an elongated (b) condition. (See Appendix for explanation of abbreviations.) The relative position of the lobes can be altered by retraction of the body resulting in pushing the hindgut rectal lobe further apically than the pericardial lobe, thus distorting the apparent length relationship. No intrinsic significance can be attached to this variation. The major variation in appearance and size is change in length of the pallial cavity. This could be measured in 35 taxa and usually is expressed to the nearest one-eighth whorl. Flattened length measure- ments are subject to greater inaccuracy and give a less functionally correlated indication of length. The most frequent states were between five-eighths and three- fourths of a whorl, found in a total of 19 taxa (fig. 50a). Variation within a population was tested on a sample of Libera f. fratercula from Motutapu Island PATTERNS OF MORPHOLOGICAL VARIATION 87 off Rarotonga (FMNH 152742). Whorl counts of the shells ranged from 6 1 A to 7, with pallial cavity length varying from five-eighths to slightly less than a full whorl. Part of the variation was an artifact of contraction. Specimens deeply retracted into the shell had a cavity length at or near the high range of the variation. By pulling back up to a quarter whorl from the aperture, the linear length needed to occupy a full whorl was reduced. This makes the increased length an artifact of spiral distance rather than a change in actual measurement. This artifact in the data prevented meaningful statistical treatment. The median pallial cavity length in the Libera sample was three-fourths whorl. Allowing for this variability, the significant depar- tures from the average pallial cavity length in adult specimens were comparatively few. Planudonta con- cava had a three-eighths whorl-long cavity, despite the shell reaching a 6% whorl count, which is significantly over the family median of 5'/2 + . Presumably this involved no proportionate change in pallial cavity length, despite the raised whorl count. A number of species had one-half whorl pallial cavities Minidonta hendersoni, Mautodontha zimmermani, Cookeconcha jugosus, Opanara areaensis densa, Thaumatodon spirrhymatum, T. hystricelloides, and Aaadonta con- stricta. These agree in that the dissected specimens have 5 to 5% whorls, with a mean of 5Vi + . This compares with a mean of 5%+ for those species with average pallial cavity lengths. Elongation of the pallial cavity (fig. 50b) is to essentially one whorl in Taipidon semimarsupialis, Nesodiscus fictus, and Endodonta fricki; IVs whorls in Aaadonta f. fuscozonata', and between 1 1 A and l'/2 whorls in Libera micrasoma and L. b. bursatella. These are either brood-chamber taxa or the species with the highest whorl counts in their genus, averaging one-fourth (Aaadonta) to seven- eighths (Nesodiscus) whorl more than any other species in their genera. It is reasonable to look upon these species as showing proportionate elongation of the pallial cavity as the whorl count increased. The dissected specimens of these taxa had a mean whorl count of 6% -, substantially above the other groups. The general elongation of the cavity correlates with the extremely shortened rectal lobes seen in Aaadonta fuscozonata (fig. 199f), Taipidon semimarsupialis (fig. 139c), and Endodonta fricki (fig. 163e). It is not the only correlative, since in Nesophila tiara (fig. 165c) the rectal lobe is virtually absent, possibly as a result of the dramatic overall size increment to over 11 mm. shell diameter in this species. The much larger whorl cross-section would provide adequate kidney volume space along the upper palatal wall without the thickness of the lobes seen in Libera (fig. 171d). Basically the variations in kidney size, shape, and pallial cavity length should be looked on as represent- ing space compromises. Presumably there is an effective kidney-volume/body-volume ratio that has to be maintained. The changes in shell growth that alter the palatal wall barrier configurations result in adjustment as to the kidney shape and lobe lengths. The available material did not lend itself to comparing foot length, shell whorl count, and pallial cavity length to see if the increase of length in the pallial cavity might be required to provide extra space for foot withdrawal. There is no question but that most of the length increase comes through extension of the relative area occupied by the respiratory surface between the anterior tip of the kidney and the pneumostome (see, for example, fig. 199a, f comparing one-half whorl and \Vs whorl pallial roofs, respectively). Other features in the pallial complex show no significant variations in those taxa dissected. The heart (H) generally is slightly more than half the length of the kidney (K) and lies almost exactly parallel to the hindgut (HG). The hindgut extends only slightly apicad of the kidney base before departing from the parietal-palatal margin (fig. 163d). It then shifts onto the palatal wall and continues apically past the stomach expansion. In most dissected specimens there are at least traces of the mantle retractor muscle (fig. 199a, MD) present, but this feature has been omitted from most drawings. In summary, the pallial complex in the Endodon- tidae is quite uniform in structure and differs significantly from that of the Charopidae in its lack of any secondary ureter. Changes in pallial cavity length correlate partly with whorl count alterations. Changes in the kidney shape and proportions correlate with changes in the upper palatal wall proportions. While the pallial complex has high information content as to family affinity, it yields little information on relation- ships within the family. Digestive system Descriptions and figures (figs. 163, 164) of Endo- donta fricki serve to illustrate the basic structures of the digestive tract and associated glands. The buccal mass (fig. 164e, B) is uniform in shape throughout the family and has a very small generative sac visible posteriorly. Often the buccal ganglion (BGN) remains attached after dissecting out the buccal mass and muscles. The two salivary glands (OG) touch above the esophagus (BE) but are not united. Their ducts (OGD) enter the top of the buccal mass on either side of the esophagus. The buccal retractor muscles (BR) insert in a U-shaped fan on the buccal mass base. The esophagus is another "space saving" organ, since it passes from the buccal mass to past the pallial cavity apex, preserving the space function of the pallial cavity for retraction of the head and foot. Expansion of the digestive tube into a stomach (fig. 164f, IZ) starts perhaps one-eighth whorl above the pallial cavity apex and is completed one-eighth whorl later. The stomach extends from one-half to more than one whorl apically, then reflects into the narrow intestine (I) that passes forward into the typical complicated looping pattern between stomach and pallial cavity apex that is characteristic of pulmonates. One loop of the intestine f FIG. 51. Radular teeth of Libera fratercula rarotongensis. Station R15, east of Avarua, Rarotonga, Cook Islands. FMNH 152744: a, central and early laterals viewed from low posterior angle at 1,850 X; b, anterior view of early laterals at 4,225 X; c, inside view of right laterals at 5,150x ; d, outside view of right laterals at 5,375x showing interrow support; e, lateromarginal transition viewed from left posterior at 2,150x ; f, detail of marginal teeth at 5,325 x. 88 FIG. 52. Radular teeth of Taipidon petricola decora. North side of Vaituha Valley, 600 ft. elevation, Eiao, Marquesas. BMNH: a, partial row of teeth viewed from anterior front at 1.275X; b, central and first lateral teeth viewed from anterior side at 3,075 X; c, worn mid-marginal teeth from right side of radula at 5,350 x ; d, worn lateral teeth from right side of radula at 3,435 x ; e, late marginal teeth at 5,215 X . 89 90 SOLEM: ENDODONTOID LAND SNAILS projects forward under the kidney base, indenting rather deeply the inner surface of the kidney. The position of the hindgut has been described previously. Variations in esophagus length correlate exactly with pallial cavity length. Variations in stomach length are a factor of how many whorls of the shell are actually occupied by the soft parts as adults (pp. 94- 95), plus changes in the cross-sectional areas of the whorls. With elongation of the soft parts and narrow- ing of the whorls, the stomach will be longer. Shortening of the soft parts and increase in the cross- sectional whorl area will shorten the stomach. Ex- treme contraction of the animal when the stomach is empty will give the illusion of a shorter stomach, since the severely compacted soft parts will collapse the early section of the stomach to permit maximum withdrawal. Only in the radula and jaw are there significant changes in size and structure. Almost all of this work was done prior to the late 1960's general availability of the scanning electron microscope. Hence the data presented here is far less satisfactory than could be accomplished with the same material today. Because the radula is extremely tiny and the individual teeth are in the 6-16 fj. size range, optical observation was quite difficult. It was not possible to observe outer marginal teeth in the vast majority of species due to mounting problems. More than in any other group of snails that I've studied, the marginal teeth of the radula fold under or fragment off during handling. With the size of the central tooth ranging from 4 X 6/1 in Taipidon centadentata to 13 X 16 JLI in Ruatara oparica reductidenta, observations on other than the point of transition to marginal teeth (indicated by endoconal appearance and tooth-size reduction) and basic cusp patterns were not possible. Recently I have been able to study the radular teeth of Thaumatodon spirrhymatum and Priceconcha tuvuthaensis (Solem, 1973d), Libera fratercula rarotongensis (fig. 51), Taipidon petricola decora (fig. 52), Endodonta fricki (fig. 53), and Thaumatodon hystricelloides (fig. 54) using the scanning electron microscope. This has added considerable information. The basic pattern is that of a tricuspid central tooth (figs. 51a; 52b) with the two ectocones quite small compared with the mesocone, and the mesocone itself often slightly shorter than the mesocone on the flanking first lateral teeth. On either side are generally five to six bicuspid laterals (figs. 51b, c) in which the ectocone is slightly less than half to only one-third the mesoconal length. There is no trace of an endocone, except in teratological rows (fig. 54b). The transition to marginal teeth generally occurs over a three-tooth spread, involving size reduction of the tooth, shorten- ing of the basal plate, appearance of an endocone, partial reduction in size of the mesocone, and increase in relative ectoconal prominence. The marginal teeth, which usually were lost in preparation, tend to more or less multicusped ectocone, smaller size, widened FlG. 53. Radular teeth of Endodonta fricki (Pfeiffer). Makalea, Waianae Mts., Oahu, Hawaii. FMNH 53042: a, early marginal teeth at 1,090 X; b, lateral teeth from left side of radula at 1,200 X (note deformed row with endocone at lower area). Extracted from a dried specimen and incompletely cleaned. form, and shortened length. Since the transition from lateral to marginal teeth occurs over several teeth, observers may count them differently. Cooke (1928) reported four laterals in Endodonta lamellosa, five in E. marsupialis, and seven or nine in E. fricki, with, respectively, 16-19, 17-20, and 12-14 marginals. These differ from my observations slightly. In the few cases Fir,. 54. Radular teeth of Thaumatodon hystricelloides (Mousson). Station 19, Lake Lanuto'o, Upolu, Western Samoa. FMNH 153130: a, worn laterals from left side of radula at 1.795X ; b, lateromarginal transition zone from right side of radula showing a deformed longitudinal row (upper left) at 1,265 X; c, early unworn laterals from right posterior at 5.975X; d, early marginals from same area of radula at 6.325X; e, late marginals from right side of radula at 1.240X; f, late marginals in detail at 3,175x. Figures b, d, e, /courtesy of Engis Equipment Company demonstration of a Cambridge scanning electron microscope. 91 92 SOLEM: ENDODONTOID LAND SNAILS where complete or nearly complete marginal fields were mounted successfully, there were 9-13 marginals, the number increasing to as many as 20 in Endodonta marsupialis, for example. The number of laterals increased to seven to eight in the three Thaumatodon, Nesophila tiara, Planudonta subplanula, Taipidon fragila, and T. varidentata. Cookeconcha hystricellus, C. hystrix, and Planudonta concava had 9-10 laterals; Cookeconcha jugosus and Taipidon semimarsupialis had 11-14 laterals; and in Taipidon centadentata there were 22-23 laterals. The increase in laterals in T. semimarsupialis was accompanied by an increase in marginals to 16-17 and a decrease in size of the teeth, as measured by the central tooth, to only 4 X 6 n, compared with the 8 X 10 ju in most other Taipidon. This change in number of lateral teeth is not size correlated, but is phyletically limited to Cookeconcha and the Taipidon- Planudonta sequence. The number of tooth rows ranged from 80 to 115 in the few species for which this could be tallied. The total number of radular teeth varied from perhaps a low of 2,700 in many Opanara to a maximum of 8,100 in T. semimarsupialis. Most of the species have about 3,000-4,000 in denticles. This is far below the typical pattern seen in zonitoid taxa such as the Pacific Island Microcystinae, Helicarioninae, Euconulinae, and Trochomorphidae, where only 8 of 169 taxa for which data are available in H. B. Baker (1938b, 1940, 1941) had radular tooth counts of under 6,000. Mean tooth counts for these family level taxa were, respectively, 14,900, 28,500, 13,450, and 11,750 teeth per radula. In the New Zealand typical Charopidae, using data from Suter (1913), total tooth numbers average only 3,175, while the modified "flammulinids" average 5,676, although ranging from 2,300 to 9,380 teeth. The possible significance of this is discussed below (pp. 104- 105). The approximate size of the central tooth was measured for each species using an optical micrometer and a Leitz Ortholux microscope with phase-contrast illumination. In Cookeconcha and Orangia, for ex- ample, the teeth were narrow and elongated, measur- ing, respectively, 6-8 X 14 n and 6-8 X 10-13 /i. The largest teeth were seen in Endodonta (13 ju square) and Ruatara (13 X 14-16 /x). The smallest were in Opanara fosbergi (5 X 8 n) and Taipidon centaden- tata (4x6 ju). None of these variations show obvious shell-size correlation. Phyletic lineages are not in- volved as Cookeconcha has elongate denticles and those of its descendent genus Endodonta are virtually square. On Rapa, Orangia has elongated denticles; those of Ruatara are very large and squarish; and Opanara has variable-sized denticles. Optical study of the Rapan species suggested yet another pattern of variation. In Opanara, Ruatara, and Orangia, by the fourth lateral there is a marked inward curve of the mesocone. By the eighth or ninth tooth from the center, there is a distinct endocone and the mesocone is like that of the typical species. Previously it was much larger. After the sixth or seventh tooth, the size decreases rapidly to typical multicuspid marginals. In Rhysoconcha, the marginals appear to be the same size as in the other taxa, but the central (6 X 8 n) and laterals appear noticeably smaller than in most other Rapan species. Unfortu- nately, their size was so small and the mounts so poor that I am uncertain as to the exact cusp structure in comparison with the larger species. The decrease in size for mid-radular teeth in Rhysoconcha, elongation in Orangia, variability in Opanara, and large square shape in Ruatara possibly indicate specialization in feeding or substrate factors. The above material had been mounted for optical study and returned before I had access to a scanning electron microscope. It was not practical to reopen this phase of the study. Scanning electron microscope observations have been made on six species of Endodontidae (Solem, 1973d; this paper), plus 39 species of Charopidae and Punctidae from Australia, New Zealand, New Cal- edonia, and Lord Howe Island. Data about the Endodontidae are summarized first. As is typical of pulmonate radulae, the central and lateral teeth show an interlocking device between longitudinal teeth rows. The stressed cusp will receive support from the basal plate in the next anterior row. The existence of this inter-row support mechanism was first reported in a variety of taxa (Solem, 1972a) and the pattern of change in the Charopidae reviewed subsequently (Solem, 1973a, pp. 166-167). The basal plate of the central tooth in the Endodontidae, as exemplified by Libera fratercula, has a raised lateral ridge on each side of the basal plate (fig. 51a). In the early laterals (figs. 51c, d) the flared lateral ridge is restricted to the ectoconal (outer) side of the tooth and functions to prevent lateral shifting of the teeth under stress. The endoconal (inner) side (fig. 51a, upper right) of the laterals has virtually no trace of the support ridge. Exactly the same pattern is seen in Taipidon (fig. 52d), Thaumatodon (fig. 54b; Solem, 1973d, figs. 6, 7), and Priceconcha (Solem, 1973d, figs. 13, 14). The marginal teeth are characterized by basal plates that are greatly reduced in length (but not width), development of a medium to prominent endocone, reduced mesocone, and often split ectocone. The transitional area from laterals to marginals (figs. 51e; 52a; 54b; Solem, 1973d, fig. 8) is short. Apparently there are significant differences in the pattern of marginal tooth structure, but because of their small size and the nature of these differences, only the few species studied with the SEM can be discussed. In Taipidon (figs. 52c, e) the cusp sits very low on the basal plate, all cusps become narrow and elongated, while the ectocone tends to become bicuspid. In Libera (figs. 51e, f) the early marginal teeth sit much higher on the basal plate, the ectocone tends more to fragment, and the mesocone is not nearly so narrow. PATTERNS OF MORPHOLOGICAL VARIATION 93 Thaumatodon hystricelloides (figs. 54d, e, f) has a very long and dagger-shaped mesocone, a slender and much shorter endocone, and initially a simple ectocone that becomes highly fragmented on outer teeth (f). The cusps of this species are elevated very high above the extremely short basal plate. Only the first few marginals were seen in T. spirrhymatum (Solem, 1973d, fig. 15), but these agree with the other Thaumatodon. In Priceconcha tuvuthaensis (Solem, 1973d, figs. 8, 9) there is quite a different pattern of ectoconal splitting on the marginals. Instead of roughly coequal splitting of the ectocone, lateral knob- like buds appear. The differences lie in the elevation of the cusps above the basal plate, the method of splitting for the ectocone, and the pattern of cusps on the marginals. While these can be studied with the SEM in angled view, they are below the level of optical examination. In summary, the Endodontidae have a relatively uniform pattern of tricuspid central, bicuspid laterals that number five to eight, only rarely increasing in number, and somewhat more numerous tricuspid to multicuspid marginal teeth. Tooth size and shape varies within lineages, with changes in tooth numbers basically occurring only in Cookeconcha and the Marquesan Taipidon-Planudonta lineage. The latter group shows the greatest amount of radular change, with Planudonta concava, Taipidon semimarsupialis, and T. centadentata showing progressively increased tooth number, but progressively decreased tooth size (table LV). The changes are not size correlated, since the teeth of Nesophila tiara, whose mean shell diameter of 11.29 mm. is vastly larger than the shell of Cookeconcha jugosus (5.01 mm.) or C. hystricellus (4.98 mm.), show no increase either in size or actual numbers. The typical radula found in the Charopidae presents a number of obvious differences. In the vast majority of species there is a tricuspid central tooth that is markedly smaller than the flanking laterals, which normally are tricuspid with equal-sized endocone and ectocone. Generally, all laterals have a narrow mesocone and large side cusps, but a few taxa show significant modifications (based on unpublished SEM observations). The New Zealand Allodiscus dimorphus (Pfeiffer, 1853) has bicuspid laterals, but typically multicuspid marginals; the New Zealand Thalassohelix propinqua (Hutton, 1883) has bicuspid laterals and unicuspid marginals; the New Zealand Serpho kiwi (Gray, 1843) has a unicuspid central, bicuspid laterals, and early marginals that approach the helicarionid marginal structure; while on Lord Howe Island the succineiform Mystivagor mastersi (Brazier, 1876) (Solem, 1973a, fig. 6) has a peculiar anterior supporting flare developed and Pseudocha- ropa lidgbirdi (Etheridge, 1889) has unicuspid margin- als. Apparently all of these modified taxa are partly- to-completely arboreal in habitat, which probably explains the selective pressure behind the modifications. The marginals in the Chaopidae vary widely in shape and form, much more than in the Endodontidae. Consideration of this variation is deferred. Stand- ing in extreme contrast to both radular types are the denticles in the Punctidae. They have a unique pattern of very tiny lateromarginal undifferentiated teeth in which there is a narrow, bicuspid tooth with slender basal plate and evenly curved anterior that rises to the cusps, which point essentially directly forward and have extremely tiny accessory cusps on each side of the tooth and then between the two main cusps. This peculiar pattern was detected by H. B. Baker many years ago (Pilsbry, 1948, p. 642, fig. 349, TABLE LV. - RADULAR TOOTH SIZE AND NUMBERS IN NESOPHILA, COOKECOHCHA, TAIPIDON, AND PLANUDONTA Species Cookeconcha hystricellus Hesophila tiara Taipidon p_. petricola varidentata fragila semimarsupialis centadentata Planudonta Tooth numbers Laterals Marginals 9-10 lit subplanula concava 6 1 7-8 13-15 22-23 10-11 10-11 8+4+ 8+++ 13 9-10 10-13 16-17 10 7+++ Central tooth size in microns 8 X 13-1 1 * 6 X lit 9-10 x lU 7-8 x 8 8 X 10 8 X 10-11 6-8 x 8 l* X 6 6 X 8-10 8 x 10 94 SOLEM: ENDODONTOID LAND SNAILS d), but is much more easily interpreted with scanning electron microscope observations. In typically unmodified endodontoid taxa, in- spection of the radular lateral teeth is quite sufficient to establish family affinities. They are bicuspid in the Endodontidae, bicuspid but very differently curved and with accessory microcusps in the Punctidae, and tricuspid in the Charopidae. This presents one of the clearest diagnostic features in terrestrial species, but is subject to at least partial convergence in the arboreal Charopidae. TABLE LVI. - SHELL SIZE AND JAW STRUCTURE IN THE ENDODONTIDAE State Separated square plates Separated elongated plates Central plates partly fused All plates fused . .' er of taxa 1 16 9 1 Shell diameter in mm. 2.93 3.TU0.19 (2.5U-5.79) 5.750.50 (U. 23-8. 99) 11.29 Jaw structure in the Endodontidae varies with size (table LVI). Typically (fig. 125g), the jaw consists of many separate, elongated chitinous plates. This is the pattern found in Opanara, Australdonta, Thau- matodon, most Aaadonta, Planudonta concava, Taipidon varidentata, and T. fragila. One species, Aaadonta fuscozonata, has the plates nearly square in shape. In a number of species, all Cookeconcha, Endodonta, Nesodiscus, Planudonta subplanula, Taipidon centadentata, and T. semimarsupialis, the central plates of the jaw are at least partly fused, although the plates on either side are clearly separated from each other and retain the typical elongated shape. Finally, in Nesophila tiara the jaw plates are completely fused together. The correlation of increas- ing jaw plate fusion with increasing size is obvious (table LVI). In the Charopidae there is a basically similar pattern of separated and elongated plates, with partial fusion occurring first in the center of the jaw, while in the Punctidae only the pattern of small, clearly separated plates has been observed. Presum- ably, this correlates with the small size of most punctids. Variations in the jaw structure, stomach length, esophageal length, and possibly the number of margin- al teeth correlate with size and whorl count factors. Radular cusp size and shape variations cannot be studied effectively by optical viewing, but only a few taxa could be examined with the scanning electron microscope. Free muscle system In all the dissected species, the right ommatopho- ral retractor passed through the penioviducal angle, while the right rhinophoral retractor passed outside the angle. Fusion of the tentacular retractors, buccal retractors and tail fan to form the columellar retractor occurs at slightly different relative positions, correlat- ing mainly with the length of the pallial cavity. The pattern of free muscle fusions detailed for Endodonta fricki and Nesophila tiara are typical. For most taxa, only obvious differences from this pattern have been annotated in the text. The origin of the columellar retractor muscle has been discussed above. Nervous system In both Thaumatodon hys trice lloides and Libera fratercula fratercula the penis is enervated from the right cerebral ganglion. The ganglia in the dissected Endodontidae are proportionately much smaller than those in such families as the Tornatellinidae and are much more heavily encased in sheets of connective tissue. Because of the limited material and reduced prominence of the structures, no attempt was made to study details of the nervous system. External body features Throughout the Endodontidae, the body color is a pale yellow white, with the eye spots and early portion of the ommatophoral retractors providing the only touch of darker color. This contrasts immediately with those Charopidae living in arboreal or semi-arboreal habitats. The tentacles, head, neck, and often the tail of these species have scattered to heavy greyish pig- mentation, although the ground strata species have the same yellow-white body color seen in the Endodontidae. The foot in the Endodontidae is universally long and slender, bluntly rounded posteriorly and truncated anteriorly, without longitudinal or transverse grooving. There is a prominent pedal and noticeably weaker suprapedal groove (FS) on the sides of the foot that unite above the tail, but there is no development of a caudal foss or caudal horn (fig. 163a). The slime network is weakly defined. Without exception the gonopore is a short vertical slit located below the right ommatophore and both above and slightly behind the right rhinophore. The mantle collar (fig. 163b) is without developed lobes or exterior protrusions, al- though several species show an extension of glandular materials onto the pallial cavity roof (pp. 84-85). In the Pacific Island Charopidae the above descriptions apply, but in the extralimital taxa there are major variations. Several New Caledonian taxa, for example, develop a "pseudo-operculum" on the tail (Solem, unpublished], and in many arboreal and semi-arboreal taxa from New Zealand there is a weak to very prominent caudal horn developed (see Climo, 1969a for references and pp. 105-106 of this monograph for a review of the controversy concerning the systematic value of this feature). Patterns of elongation As outlined below (pp. 113-114), one of the repetitive trends within the Endodontidae is for whorl count and size increase to the "Nesodiscus" and "brood chamber" levels of specialization. While this involves some elongation of the soft parts, a much PATTERNS OF MORPHOLOGICAL VARIATION 95 more typical pattern is for the animal to withdraw in part from the upper whorl of the shells. In groups such as the land prosobranch family Pomatiasidae, the pulmonate Urocoptidae, and such subulinids as Ru- mina decollata (Linne), the early whorls will be evacuated by the animal which seals off the upper whorls by a thick calcareous plug. The early whorls usually break off. This reduces shell length highly effectively with cylindrical coiling patterns, but is not an option open to planulate or heliciform taxa. Where these species show a great increase in whorl count over "typical" taxa, such as in the trochomorphid Coxia m. macgregori (Cox, 1870), almost half of the 10% whorls of the shell are above the apex of the soft parts (fig. 55a). In Libera fratercula (fig. 55b), a seven-whorled shell may have only the lower 3V4 whorls occupied by the animal. As outlined elsewhere (Solem, 1969b), the upper whorls are filled with calcium carbonate by the snail and thus provide a source of shell calcium for the young that hatch in the brood chamber, then eat their way out through the shell apex. So far as is known, this is the only species in the Endodontidae to fill in the apical whorls with calcium, but in most of the species a part to all of at least the nuclear whorls are not occupied by the adult animal. Because only a few species were available with shell and soft parts still together, I can present no statistical data concerning the variations in withdrawal from the upper whorls. Coxia macgregori was chosen for comparison because it shows the greatest whorl count of any aulacopod land snail species available to me with soft parts. A member of the limacacean family Trochomorphidae, it obviously differs in having a typically sigmurethrous ureter (KD) and in numerous genital and radular features, but it is comparable in terms of elongation patterns. The pallial cavity is about 1% whorls long, the stomach occupies seven- eighths whorl after a space for the albumen gland and intestinal loops, the ovotestis (G) is strung along almost five-eights whorl, and the digestive gland extends 2 3 /4 whorls above the ovotestis apex. The upper 5Vs whorls of the shell are "empty space." In Libera the pallial cavity extends about seven-eighths whorl, followed by a short gap for the albumen gland and intestinal loops, with the stomach occupying a full one-half whorl, the ovotestis a little less than one-half whorl, and the digestive tissue extending nearly l'/4 more whorls. A total of slightly over 3V4 whorls in the seven-whorl shell is "adult occupied" in Libera, compared with the 5% of 10% whorls in Coxia. The details of organ lengths are different, but the general pattern of withdrawal from apical whorls is equivalent. This is one way of partly coping with increased visceral hump length. Yet another way is through differential elongation of organs. The counterpart of visceral hump elongation in many whorled shells is visceral hump shortening in "semi-slugs," where organs must be compacted, rather than elongated. In both situations the change seems to occur in morphological KX IZ HG KD FIG. 55. Animal length and shell whorl count in elongated taxa: a, Coxia; b, Libera f. fratercula. Identified structures are: A - anus; G - ovotestis; GD - hermaphroditic duct; GG - albumen gland; HG - hindgut; IZ - stomach; K - kidney; KD - ureter; KX - ureteric pore; MC - mantle collar; Z - digestive gland. zones that encompass all the organ systems that pass through the zone. If the pallial cavity is elongated, for example, the pallial gonoducts will lengthen, whereas if the head and neck are elongated (or compacted) it is the terminal genitalia whose proportions will be shifted. The most frequent example of this alteration in the genitalia of the Endodontidae, and the easiest to quantify, is the change in relative lengths of the free oviduct and prostate. For 43 taxa it was possible to score this feature, with 10 taxa (three Cookeconcha, two Orangia, Nesophila tiara, Taipidon centadentata, T. varidentata, Thaumatodon hystricelloides, and T. euaensis) having the free oviduct distinctly shorter than the prostate, 14 taxa (two Ruatara, Opanara altiapica, three Taipidon, two Planudonta, Libera cookeana, four Aaadonta, and Thaumatodon spirrhymatum) having them about the same length, 96 SOLEM: ENDODONTOID LAND SNAILS 15- 14- 13- 12- 11- 10- 9- 8- 7- 6- 5- 4- 3- 2- 1- *** * T I 6 7 8 Shell Diameter I 10 I 11 I 12 I 13 FIG. 56. Penis length and shell diameter in the Endodontidae. Those species with columellar muscle origin of the penial retractor are indicated by "dots," those with diaphragm origin by "stars." and 19 taxa (one Minidonta, three Opanara, one Rhysoconcha, Orangia maituatensis, two Austral- donta, Taipidon fragila, Nesodiscus fictus, two Endo- donta, five Libera, Aaadonta fuscozonata, and Price- concha) having the free oviduct much longer than the prostate. Species with these three states do not differ significantly as groups in either shell size or whorl count. The small Minidonta and Rhysoconcha, for example, which have about five whorls, have the long free oviduct, and yet Orangia cookei has a short free oviduct. If there is a shift within a genus or between derivative genera from short to equal, or equal to long, then there is a significant (half whorl or more) increase in whorl count. TABLE LVII. - PATTERNS OF AULACOPOD RADULAR DENTITION Family taxon Total teeth Number of - X rows Lateral teeth Marginal teeth Endodontidae 3,550 (2,850-1+, 095) 98.it (6, 8, 37, 22) Charopidae 3,76? (2,100-9,380) 102.7 (15, 15, 58, 58) Microcystinae 11,750 (3,850-31,800) 98.3 (111, 127, 129, 1^3) Euconulinae 12,875 (U, 550-36, 900) 96.0 (17, 19, 18, 20) Helicarioninae 26,800 (3,900-63,000) 107.1 (13, ll+, 16, 17) Trochomorphinae lU,150 (1+, 700-25,600) 121+. 5 IQ.I (5-17) (28, 28, 31, 31) 7.2 (5-23) 7.1 (3-18) 8.0 (2-13) 8.U (1-12) 16.9 (5-38) 103.1 (30-252) 1+3.8 (19-70) 11.2 (7-19) 1^. 5 (6-32) 1+8.3 (21+-175) 5l+. 7 (19-137) TABLE LVIII. - MAJOR DIFFERENCES BETWEEN ENDODONTIDAE AND CHAROPIDAE Character Ovotestis Prostate-uterus Epiphallus Verge or vergic papilla Terminal female organs Kidney Secondary ureter Radular laterals State in - Endodontidae many follicles in line along duct ; angled to shell axis s epar at e due t s usually absent , penis chamber derived absent simple tubes weakly bilobed absent bicuspid Shell apical sculpture radials usually dominant, microspirals "squiggly" Apertural denticles present; microdenticles uniformly triangular Shell sculpture formation mostly in calcareous layers Charopidae few follicles in usually one or two clumps ; curved around axis fused with a common lumen usually present, vas deferens derived often present usually internal complex structures strongly bilobed present almost always tricuspid spirals usually dominant in Pacific Island taxa absent in most , when present microdenticles variable mostly in periostracal layers 97 98 SOLEM: ENDODONTOID LAND SNAILS TABLE LIX. - MINOR DIFFERENCES BETWEEN ENDODONTIDAE AND CHAROPIDAE Character State in - Talon head Albumen gland Prostate Uterus Penis pilasters Penial retractor insertion Endodontidae elongately oval longer, rarely indented by intestine, alveoli smaller alveoli larger, shorter, in rows less sharply differentiated two longitudinal penis apex or side Spermathecal insertion Mantle glands onto pallial roof Radular central variable often slightly smaller than 1st lateral Charopidae globose shorter, usually indented by intestine, alveoli larger alveoli longer, slenderer, irregular spacing sharply differentiated highly variable, but not two longitudinal penis-epiphallus junction, epiphallus, or penis with vas deferens piercing muscle free oviduct rarely much smaller than 1st lateral This is a simple space problem solution. The relatively featureless free oviduct and vas deferens sections of the genitalia tend to lengthen more than do the prostate and uterus, which are more complex in structure. The penis, with its retractor muscle, generally extends from the atrium to the pallial cavity apex. Variations in the proportions between penis length and penial retractor muscle length were not studied. Since the penis length was recorded, it is possible to plot it against shell diameter for either the dissected individ- uals or the mean diameter for that species. The results are presented in Figure 56; those species in which the penial retractor muscle origin has shifted to the columellar muscle are indicated by "dots," while those with diaphragm origin are indicated by "stars." The greater penial length obtainable with the shift to the columellar muscle origin is obvious. Particularly since the penis can vary in length because of interactions with sympatric congeners (p. 82), the close correlation between shell diameter and penis length is quite remarkable. Their correlation coefficient is 0.82 when the columellar insertion taxa are excluded, and 0.74 even if they are included. Thus the departure of penis length from the plotted regression line in Figure 56 could be used as a quick estimate to see whether species interactions were involved as a selective factor in the penis length of an endodontid, when compared with others in its genus. The taxa furthest from the regression line to the right are the two Nesodiscus. Why they should have such a relatively short penis is unknown. Summary of anatomical variation The Endodontidae have a basically conservative body plan that offers a number of contrasts to the structures seen in the Charopidae. These differences are summarized in Tables LVIII and LIX as part of the family level classification discussion. Many fea- tures of the anatomy in the Endodontidae penis length, degree of jaw fusion, stomach length, esopha- geal length, length of hermaphroditic duct, proportion- ate lengths of free oviduct and prostate vary in direct relationship to whorl count and body size. Changes in organ position and shapes in the pallial complex relate to shell barriers and shell shape features as space accommodations, while partial coiling of the hermaphroditic duct may serve as an indicator of whorl count reduction from the condition found in the immediate ancestors. Other features, particularly involving the penis pilaster pattern, seem to be involved in species recognition among sympatric congeners and have great utility to the taxonomist in sorting out "sibling species." Both penis length and possibly the length of the talon shaft also may be involved in species recognition interactions. Very limited data suggests that possibly the size and shape of radular central and lateral teeth may be involved in niche specialization, PATTERNS OF MORPHOLOGICAL VARIATION 99 but the work needed to test this hypothesis could not be undertaken. There are comparatively few clear progressive trends in anatomical variation or addition of new structures, which stands in very marked contrast to the situation in the Charopidae. There is a change in pilaster shape from low and rounded to high and lamellate. Several taxa show a tendency toward adding glandular tissue between the penis head and penial retractor muscle, thus giving a potential for "epiphal- lic" secretions. One group, the Thaumatodon-Aa- adonta complex, has the penial retractor shifted to the side of the penis, a much altered entrance of the vas deferens, and thus an equivalent to a "penial epi- phallus" has developed. In Taipidon and Planudonta, the only Marquesan endodontids, there is addition of a pustulate zone on the penis interior that has no equivalent in other genera. In two separate areas, involving the Hawaiian Cookeconcha and the two Marquesan genera, there is an expansion in radular tooth numbers. Other features, such as the origin of the penial retractor muscle (diaphragm or columellar muscle) and spermathecal insertion (on penis, atrium, or free oviduct), vary without direct size correlation and the changes occurred in several lineages. The shift of the penial retractor muscle to the columellar muscle does permit a definite increase in penis length, but there is no indication of any size "trigger" to this shift. Probably it occurred as a rare mutation, with the resultant opportunity for penial enlargement providing the size-release mechanism. HABITAT RANGE AND EXTINCTION With few exceptions, data on the observed niche for endodontids are monotonous. Found under stones, in talus slopes, in or under rotting logs, in leaf litter in heavy forest all summarize the classic pattern for litter-and-leaf-mould forest dwellers. Essentially all localities involve primary forest situations. The endodontids were ground stratum inhabitants of "primary forests" on the Pacific Islands. There are only a handful of exceptions. Pilsbry and Vanatta (1906, p. 783) reported that species of Cookeconcha "live on dead stumps and logs, and under the bark of dead trees, but also among fallen leaves." They also have been found in heavy moss on large boulders and at low levels on tree trunks (Solem, personal observation). Libera b. bursatella was taken in the axils of Freycinetia at 4,700-5,500 ft. elevation on Mt. Aorai, Tahiti. The Lau Archipelago Price- concha tuvuthaensis Solem (1973d, p. 24) was taken on tree trunks up to 10 ft. above ground level. Libera fratercula lives in coastal forests on several islands of the Cook group and has become adapted to living under coral rock in the narrow shore zone of storm- tossed boulders (pp. 418-419). Material of Rhysoconcha was collected by members of the Mangarevan Ex- pedition from coffee plantations, native forest, and mixed vegetation areas in the Maitua region on Rapa. The resulting apparently hybrid populations of Rhyso- concha from ecotonal stations present a highly interesting phenomenon (pp. 264-265). The Cookeconcha and Libera bursatella ex- ceptions are in zones where the rainfall exceeds 175- 200 in. annually, so that water conservation selective pressure would be minimal. Why L. fratercula exists successfully in a shore zone that is subject to at least short periodic droughts is unknown. The general pattern of the endodontids being restricted to the ground level of primary forests is clear. This contrasts greatly with the Charopidae, where a large number of species are arboreal or semi-arboreal. This difference probably can be explained by the difference in excretion regimes. The Charopidae possess a water- conserving secondary ureter, while the Endodontidae do not. They must periodically use part of the pallial water reservoir to evacuate excreted matter, and thus are more tightly tied to the high humidity levels of the primary forest ground stratum. In Australia, New Zealand, New Caledonia, Viti Levu, Upolu, and Tahiti, I have found charopids in non-primary forest and even on the fringes of plantations, but except for the records of Libera fratercula and Rhysoconcha, no endodontids have been taken from disturbed primary forest or secondary vegetation zones. This could relate to the change in litter composition and/ or humidity levels when the forest is opened up to sunlight drying, or a subtle alteration in food source. In Hawaii, on both Oahu and Kauai, traces of endodontids were found only in isolated high mountain patches of native plants, while on the island of Upolu, Price and I took material of Thaumatodon hystricelloides (Mousson) only at relatively high elevations in heavy forest that had not been invaded by introduced ants (see Wilson and Taylor, 1967 for an account of introduced Polynesian ants). In the 1860's, T. hystricelloides was common in lowland forests, but today it is restricted in its distribution. This correlates with the presence or absence of introduced ants, particularly the rapacious Pheidole megacephala. On both Oahu and Upolu, I have observed the concordance between the presence of swarming ants and the absence of many endemic snails and insects. This was documented in detail by Zimmerman (1948, pp. 172-177) for Hawaii. The possibility that this exists in relation to the endodontid fauna of Rapa, for example, has considerable evidence in its favor. Wilson and Taylor (1967, p. 6, table 2), record eight species of introduced ants from Rapa. From September 6, 1963 through December 15, 1963, J. L. Gates Clarke of the National Museum of Natural History collected insects and some land snails on Rapa (Clarke, 1971, pp. 1-26). Through the kindness of Dr. Joseph Rosewater, it was possible to examine this material. Endemic tornatellin- ids and zonitids were represented, but there were no endodontids. Subsequently, Dr. Harald Rehder of the National Museum of Natural History visited Rapa to collect marine mollusks. He is also an experienced land snail collector and, at my request, made a special trip to the Maitua area and searched, without success, for endodontids. Clarke (1971, p. 10, fig. 12) illustrated yet another disturbing factor in the ecology, the presence of goats on even some of the very steepest slopes. The ability of goats to seriously alter ground strata environments is legendary. The reason for the apparent endodontid absence wherever ants are common probably relates not to adult predation, since the apertural barriers of endodontids would presumably be effective against ant 100 HABITAT RANGE AND EXTINCTION 101 predation, but to egg or juvenile predation. The habit of egg deposition in the shell umbilicus common to endodontids would provide no protection against the ant mouth parts. Hence establishment of a foraging ant colony in an area could easily prevent successful reproduction of endodontids through continued loss of eggs from the umbilical cavities, even though conceiv- ably the adults might not be bothered. Between the visits to Rapa of the Mangarevan Expedition of the Bishop Museum in 1934 and Clarke's visit in 1963, "noticeable reduction of forest" occurred (Clarke, 1971, p. 9). The combination of reduction in forest cover, high-altitude disturbance by goats, and activities of ants easily could combine to produce great reduction, if not near total extinction of the endodon- tid fauna on Rapa. The Mangarevan endodontids had been wiped out by habitat alteration prior to 1934, and it may well be that the Rapan radiation has joined the ranks of the extinct. The material of Priceconcha (Solem, 1973d, p. 24, fig. 19, a) was heavily parasitized and this could indicate yet another factor limiting the current western distribution of the Endodontidae. No other endodontid or charopid specimens were seen with any trace of parasites. Habitat disturbance, introduced predatory ants, and possibly parasites acting separately or in combination would effectively explain the rapid ex- tinction of the endodontid fauna. Because of structural limitations in the pallial complex, they basically were restricted to ground strata in primary forests and thus were among the first taxa destroyed or displaced by human disturbance or the addition of ground-litter predators. Their low diversity in Western Samoa (only Thaumatodon on Upolu is known) may well stem from the presence of endemic ants in Samoa (Wilson and Taylor, 1967), but not on the Cook, Society, Austral, Marquesan, or Hawaiian Islands, where endodontids were quite abundant. PHYLOGENY AND CLASSIFICATION Before discussing the variation patterns found within the Endodontidae, proposing a phylogeny, and deriving a classification scheme from the proposed phylogeny, it is necessary to place the endodontoid snails within a broader context. The higher class- ification and phylogenetic relationships of gastropods still are controversial. Elsewhere I (Solem, 1974; In press B) have discussed the possible origin of snails and reviewed the higher level classification of land snails. These papers present several changes from the summary given in Solem (1959a, pp. 32-36), which was based on the classic accounts by Pilsbry (1900a, b) and H. B. Baker (1955). PHYLOGENETIC POSITION OF THE ENDODONTOID SNAILS I concur with Fretter and Graham (1962, p. 612) that the euthyneurous condition in the Opistho- branchia and Pulmonata are derived independently. A basic classification into the Subclasses Prosobranchia, Opisthobranchia, and Pulmonata thus reflects phylogeny far better than the split into Subclass Streptoneura ( = Prosobranchia) and Euthyneura ( = Opisthobranchia and Pulmonata) used by Taylor and Sohl (1962). Elsewhere I have reviewed the evidence that the Pulmonata are a grade containing three superorders, the Basommatophor a, Systellommatophora, and the Stylommatophora (Solem, In press B). These groups are "pulmonate" in the same sense that the monotremes, marsupials, and eutherians are "mammalian." It is thus quite possible that the "Pulmonata" is polyphyletic, but in the same way that the "Mammalia" is polyphyletic. Within the Stylommatophora, authors either list 12 superfamilies (Thiele, 1931, pp. 492-734) or recognize a series of higher categories, based on the divisions proposed and amplified by Pilsbry (1900a, b, 1918, 1948) and H. B. Baker (1955, 1963). Modification of the latter scheme (Solem, In press B) has been based on work suggested by the present study that led in turn to re-evaluating the basic trends along land snails. This work deliberately parallels the type of analysis done by Romer, Simpson, and others concerning the vertebrates. This methodology involves attempting to delineate the basic patterns of ecological advances made within a large taxonomic unit, attempting to identify the physiological factors or the structures (preadaptations) that permitted crossing ecological thresholds, and then identify the adaptations that permitted consolidating this gain through successful adaptive radiations. In the absence of any direct evidence from fossils, determining convergence in structures through analysis of ontogenetic changes is a powerful tool, since the same structure, if developed in different ways, does not suggest close phyletic relation- ship but rather equivalent life styles. This can be applied more successfully to snails than many groups, since the shell grows by edge accretion, leaving a frozen record of life stages visible even when the animal is fully adult. The basic problem of land life for a snail is water conservation. The land-dwelling prosobranchs have an open pallial cavity, are active only under conditions of very high humidity, and depend for water conservation on sealing themselves behind the operculum when retracted. The pulmonates have bonded the mantle cavity to the body, retaining a pallial cavity that is open to the exterior through the pneumostome. This greatly reduces water loss, and, in addition, permits the pallial cavity to hold a significantly large reserve supply of extrasomatic water (Blinn, 1964). In marine mollusks and land prosobranchs the kidney opens near the posterior of the pallial cavity. In the marine and fresh-water species, water currents sweep the excreted matter out of the cavity, but in land prosobranchs this option is not available. Pallial water or a "squirt" of excreted water must be used, at least occasionally, to flush out excreted matter. In the Basommatophora (Delhaye and Bouillon, 1972a) the land-dwelling Ellobiidae have a simple kidney with no ureter, while the fresh-water dwellers have an anterior nephridial pouch that is involved in osmoregulation. The whole kidney is elongated and extends well toward the anterior edge of the pallial cavity. In groups such as the Planorbidae (F. C. Baker, 1945, pis. 44-47) there is a reflexed anterior termina- tion that has been called a ureter. In the Stylommatophora there are fundamentally different structural patterns that have been used, first by Pilsbry (1900a, b), to delineate several orders. The basic configurations of the pallial complex in the ordinal groups as outlined by Pilsbry (1918, 1948) and H. B. Baker (1955, 1963) were summarized in an earlier paper of mine (Solem, 1959a, pp. 32-35, fig. 1). There are only three basic configurations among the five orders. The kidney in the Orthurethra resembles 102 PHYLOGENY AND CLASSIFICATION 103 that of the Basommatophora in that it extends far forward toward the pneumostome, tapering gradually, and ending in an anterior ureteric pore that opens inside a reflexed ridge that extends partly posteriorly (Pilsbry, 1900a, pi. XVII, fig. 3). In the Mesurethra, the kidney is shortened and triangular and remains at the pallial cavity posterior; there is no strong anterior extension of the kidney and the ureteric pore is a simple opening at the anterior kidney tip. This condition is found in the Cerionidae, Clausiliidae, Strophocheilidae, and Dorcasiidae, but not the Coril- lidae (Solem, 1966a, pp. 94-95) which originally were included in the Mesurethra. The Sigmurethra have a ureter starting at or near the anterior tip of the kidney, following its upper margin back to the posterior of the kidney, then reflexing forward along the hindgut as either an open groove (most Holopodopes) or closed tube that is heavily vascular- ized (most Aulacopoda and Holopoda). The initial backward extending part is called the "primary ureter" and the section along the hindgut the "secondary ureter." Delhaye and Bouillon (1972b) reported that the orthurethran kidney differs significantly in histo- logical structure from that of the Sigmurethra, and they propose that the Sigmurethra were derived from the Mesurethra by addition of the ureter to the mesurethran kidney. Rather than the Sigmurethra being descendants of the "more primitive" Orthu- rethra, these taxa are parallel experiments probably derived independently from the "Urpulmonata," what- ever group that may be. The above summarizes the basic structural pat- terns seen in the stylommatophoran pallial complex. On the basis of both gross morphology and histology, the orthurethran kidney is very different from that of the mesurethran and sigmurethran lineages. The derivation of sigmurethran type from the mesurethran situation has an appealing simplicity, but requires further investigation because of complicating vari- ations in several groups. The two ordinal groups with variations from these basic patterns are the Tracheo- pulmonata (Family Athoracophoridae) and Heteru- rethra (Family Succineidae). The former are slugs with the visceral hump organs compressed completely into the foot cavity. Their multi-looped ureter is a secondary modification to the visceral hump reduction. The heterurethrous pallial configuration was suggested by H. B. Baker (1955) as the probable ancestral condition to the Sigmurethra. Elsewhere I (Solem, 1969b; In press B) have reviewed the relationships of the Succineidae and suggested that they are modified Sigmurethra rather than being primitive. Bouillon and Delhaye (1970) reported that the basic structure of the kidney and ureter in the Succineidae and Sigmurethra were the same, but subsequently (Delhaye and Bouillon, 1972b, p. 141) concluded that because the opening from the kidney into the ureter differs in the Sigmurethra and Heterurethra, they are not related. Generally, they live under quite different water regimes, the Heterurethra in semi-aquatic and the Sigmurethra in terrestrial, often water shortage condi- tions. The addition of a sphincter to the kidney pore in the Sigmurethra, or its secondary loss in the Hete- rurthra, if my interpretation of their relationships is correct, is not a major difference. I consider it highly significant, in terms of judging relative phyletic position, that all land-slug taxa have sigmurethrous ureters. There is great water conservation potential in the closed, complete ureter that opens to the exterior at the pneumostome. This permits keeping the pallial water supply for replacing water evaporated from the extended head and foot. The ureter can function to resorb water from the excretory products, and no pallial water need be used to flush out the excretory products. The evolution of a "pseudosigmurethrous" pallial structure in some enids (Solem, 1964) is another point suggesting the funda- mental importance of the pallial structures to progres- sive land-snail evolution. With this background information, the pallial complex in the Endodontidae can be compared with the basic patterns. It is closest to the sigmurethrous condition, but in having only a primary ureter, with no trace of a secondary ureteric groove or tube, it represents a significantly different structure. A possi- bly parallel situation is seen in the Australian Caryodidae, where the primary ureter opens posteriorly, without there being any rectal kidney lobe, much less the slight reflexion of the ureter seen in the Endodontidae. The similarity is undoubtedly convergent, since the endodontid and caryodid ureters are very different in internal structure (Solem, unpublished). Whether the ureter in the Endodontidae is a forerunner of that seen in the Charopidae and typical Sigmurethra, or an independent experiment is uncertain. Certainly this represents a major difference in structure. The orthurethrous kidney is too different to be viewed as a potential ancestor to the endodontid condition. In terms of pallial complex configuration, the Endodontidae are less advanced than the rest of the Sigmurethra, but more advanced than the Mesu- rethran taxa. There is no evidence at all that the Mesurethra are ancestral to the endodontoid group. In respect to other organ systems, the trends of variation are either less clearly delineated, or else simply have not been analyzed in sufficient detail to permit firm phyletic statements. To summarize the limited data presented elsewhere (Solem, In press B), in general, it would be correct to say that taxa with no spermatophore formation (=no differentiated epi- phallus), no accessory dart sacs or mucus glands, no vergic structure, and separated pallial gonoducts (prostate and uterus completely separate tubes) are more "primitive" (= generalized) than those with a hard spermatophore, dart sacs and/or mucus glands, a verge, and united pallial gonoducts. These criteria are based on the assumptions that: 1) hermaphrodit- 104 SOLEM: ENDODONTOID LAND SNAILS ism in snails was achieved by combining separate male and female systems. Union of ovary and testis into an ovotestis was followed by subsequent union of progres- sively lower portions of the pallial gonoducts (Solem, 1972b, pp. 108-112); 2) transfer of sperm by snails in a "protective package" is (subject to secondary modification) more advanced than transfer of sperm loose in fluid (for mollusks, not mammals). The "advanced" conditions have been arrived at indepen- dently in each of several lineages, judged by analysis of structures in groups with different types of such accessory structures. The helicarionid, helicid, and helminthoglyptid dart sacs, for example, are very different in structure, although performing identical functions. Union of the prostate and uterus into a "spermoviduct" can be traced as separate devel- opments in at least helicid, helicarionid, endodontoid, partulid, pupillid, and ellobiid stocks (Solem, 1972b). H. B. Baker (1955, 1956, 1962a) recognized three major groups among the Sigmurethra. The Au- lacopoda and Holopoda, established by Pilsbry (1896), differ in foot structure, pedal groove presence or absence, basic radular features, and shell character- istics. The Holopodopes contains mainly elongated herbivorous and specialized carnivorous taxa that show numerous differences from the other groups. The Holopoda are universally accepted as being more advanced and, in most characters, are derivable from the Mesurethra. Relationships of the Holopodopes to the other groups are uncertain. The endodontoid snails have the basic foot and radular structures of the Aulacopoda. While the Charopidae have a sigmurethrous ureter, differentiated epiphallus, usually a verge, and fused pallial gonoducts, the Endodontidae lack a secondary ureter, only rarely have any indication of an epiphallus, lack a verge, and have separated pallial gonoducts. Thus the Endodontidae have aulacopod features, but in the few pallial and genital structures where it is possible to make any positive statements concerning primitive versus derived characters, in every case the Endodon- tidae show the primitive condition. Because of basic differences in structure, it is not possible to derive the Sigmurethra from the Orthurethra. The large and specialized Mesurethra show many genital features that are more advanced than the structures seen in the Endodontidae. We are thus left with an inability to focus on any group of living land snails as possessing a greater number of generalized structures than the Endodontidae. Questions concerning poten- tial derivation of the other endodontoid families from the Endodontidae are deferred until the second monograph. Family groupings for the more advanced endodontoids also will be considered elsewhere, except for the many comparisons with the Pacific Island Charopidae. The basic division within the Aulacopoda was recognized by Pilsbry (1896, p. 110), who characterized the superfamilies later named Limacacea (H. B. Baker, 1941, p. 206) and Arionacea (H. B. Baker, 1955, p. 109). The cited "key character" was the structure of the radular marginal teeth, but the smooth and shiny, often colorful "limacoid" shell, frequent development of highly elaborate accessory genital structures, and strong development of mantle collar lobes and exten- sions in the Limacacea, stand in great contrast to the dull, heavily sculptured, frequently flammulated shell and comparatively rare development of accessory genital structures or mantle extensions in the "endodontoid" Arionacea. Preliminary work (Solem, unpublished) suggests that some of the Austrozelandic arionaceans ( = advanced Charopidae) may be partly transitional in some characters to the shell-bearing limacaceans. The general "Gondwanaland" dominance of the arionaceans and "Laurasian" dominance of the limacaceans probably have influenced the general acceptance of the Limacacea as the derived taxon. This view very probably is correct, but presentation of the evidence must be postponed. Although the basic difference between the Ario- nacea and Limacacea is usually cited as the shape of the marginal teeth narrow, lengthened basal plates with unicuspid, multicuspid, or bicuspid teeth in the Limacacea and short, wide, often squarish basal plates with unicuspid or several cusped teeth in the Ario- nacea there also are differences in the sheer number of teeth. The data are spotty, particularly since few row counts were made during this study and my observations on the number of marginal teeth in the Endodontidae and Charopidae are quite incomplete. Nonetheless, the basic trend is clear. Data have been compiled from this report for the Endodontidae. Many tooth counts, but few row counts, are available for the New Zealand Charopidae (listed as Phenocohelicidae, Endodontidae, and Oto- concha) from Suter (1913, pp. 620-732). Massive information on the Pacific Island limacacean radulae was presented by H. B. Baker (1938b, 1940, 1941). A rough estimate of the total teeth on each radula was calculated by multiplying the individual row count by the number of rows on the radula. Counts of lateral teeth and marginal teeth in a half row for each species were averaged for major taxonomic units. These data are summarized in Table LVII. The numbers under the taxon name refer to the number of observations included in each column to the right. The clas- sification of the limacacean groups is slightly altered from H. B. Baker, in that the Trochomorphidae is listed as a full family; and the Helicarioninae includes the Sesarinae of H. B. Baker (1941, pp. 238-263) as was suggested earlier (Solem, 1966a, pp. 22-24). The low total tooth count on the radulae in the Endodontidae and Charopidae stands in great contrast to the situation in the limacacean groups. The figure for the Endodontidae omits the two Taipidon with grossly enlarged tooth counts. They are quite atypical for the family. Their inclusion in such a small sample PHYLOGENY AND CLASSIFICATION 105 would distort the results. The slightly higher marginal tooth count for the New Zealand Charopidae reflects the inclusion of the several "flammulinids" with altered marginal teeth. These are altered not only in tooth number, but also in form and cusp structure to the point that Suter (1894a, p. 62) had stated that "the radula is more or less pseudo-zonitoid" in these genera. The Microcystinae and Euconulinae, which are the most generalized and smallest sized members of their respective families, have added only one lateral tooth, but tripled to quadrupled the number of marginal teeth. This change in both number and tooth form of the marginals suggests a major shift in feeding. By use of critical point drying techniques followed by SEM observation with the radular ribbon in a normal position (see Runham, 1969, fig. 1, for an example of this technique) much information could be gathered on the differences in functioning. This could be the key to understanding the adaptive shift from the arionoids to the limacoids. While the Euconulinae and Microcystinae are comparable in adult size to the Endodontidae and Charopidae, the Pacific Island Helicarioninae (Or- piella, Dendrotrochus, Ryssota, Epiglypta, Helicarion) are 10-55 mm. in shell diameter, and thus much larger in size. Similarly, the Trochomorphidae are mostly 8- 20 mm. in shell diameter, again substantially exceeding the endodontoids in size. Hence the increased number of tooth rows in both taxa, and greatly increased tooth numbers in the Helicarioninae can be partly the result of simple size increase. This only accentuates the basic pattern in which the limacaceans are seen to differ mainly through the multiplication and change in form of the marginal teeth on the radula. A comparative study of the more generalized limacaceans and the Austrozelandic charopids with "pseudo-zonitoid" teeth might yield considerable data on the inter-relation- ships of these superfamilies. In summary of the above discussion, the en- dodontoid snails are a group that are "comfortably sigmurethran" (Charopidae) to "protosigmurethran" (Endodontidae). They comprise the least specialized complex of the Aulacopoda. This group parallels the Holopoda, but is not as probable an ancestor to the Holopoda as would be the Mesurethra. While several family groups may be derived from the endodontoid complex (including the Limacacea), no extant group of land snails can be pointed out as possibly representing the stem group for the endodontoid complex. The statement that the "Endodontidae probably are the most primitive living sigmurethrans" (Solem, 1959a, p. 77), which was based more on intuition than evidence, has not been altered by more than a decade of patient poking into endodontoid guts. What has been altered is the concept of family units and definitions expressed in the same paper. A review of family units precedes discussion of phylogeny within the Endodontidae. FAMILY CLASSIFICATION OF THE ENDODONTOIDS The following family level names are available for endodontoid snails. They are listed in order of nomenclatural priority. Punctinae Morse (1864, p. 27) Patulinae Tryon (1866, p. 243) Charopidae Hutton (1884b, p. 199) Phenacohelicidae Suter (1892a, p. 270) Otoconchinae Cockerell (1893, pp. 188, 205) Endodontidae Pilsbry, 1895 (Pilsbry, 1893-1895, p. xxviii) Flammulinidae Crosse (1894, p. 210) Thysanotinae Godwin-Austen, 1907 (Godwin-Austen, 1889-1914, p. 189) Laominae Suter (1913, p. 732) Goniodiscinae Wagner (1927, p. 305) Helicodiscinae Pilsbry in H. B. Baker (1927, pp. 226, 230) Rotadiscinae H. B. Baker (1927, pp. 226, 228) Stenopylinae Thiele (1931, p. 569) Amphidoxinae Thiele (1931, p. 575) Discinae Thiele (1931, p. 578) Dipnelicidae Iredale (1937b, pp. 22-23) Paralaomidae Iredale (1941a, p. 263) Hedleyoconchidae Iredale (1942, pp. 34-35) Pseudocharopidae Iredale (1944, p. 312) The Iredale taxa are virtually nomina nuda, and consideration of the extralimital units Thysanotinae, Goniodiscinae, Discinae, Helicodiscinae, Rotadiscinae, Amphidoxinae, and Stenopylinae is deferred. Of the remaining taxa, the name Patulinae is ignored for the following reasons. The describer (Tryon, 1866, p. 242) noted that the Patulinae was "not proposed with any intention but to facilitate the determination of species." Although used as a family name by Moellendorff (1890, p. 221; 1900, p. 109), it has been ignored by other authors of that period and by subsequent students, until listed with disapproval by H. B. Baker (1956, pp. 134, 138). It is now equivalent in modern context to the Goniodiscinae and Discinae, since the genus Patula, after a very long and checkered career, has settled as a subjective synonym of Discus. May both Patula and Patulinae rest in peace. Morse (1864, p. 27) established the Punctinae on the basis of having a jaw composed of 16 distinct plates, and minute radular teeth that he thought resembled those of Carychium, an ellobiid, under optical study. There are shell, radular, and genital features which combine to separate the Punctidae as a family unit (Solem, unpublished). The Laominae of Suter (1913) is not separable from the Punctidae, as has been recognized by Pilsbry (1893-1895), Thiele (1931), and Climo (1969a). The first use of the name Charopidae (Hutton, 1884b, p. 199) was based on the heliciform shell and development of a caudal mucus gland. The latter feature, which not only is very characteristic of arboreal snails in general, but is highly variable in degree of development at a very low taxonomic level, was the subject of more than a decade of controversy concerning endodontoid classification. In a series of papers Pilsbry (1892a, pp. 54-55; 1892b, pp. 68-69; 106 SOLEM: ENDODONTOID LAND SNAILS 1893a, pp. 401-402; 1893b), Hedley (1893a, p. 163), and Ihering (1893, p. 121) downgraded the importance of this character, although Crosse (1894, pp. 210, 219) and Moellendorff (1895, pp. 157-158; 1899; 1900, p. 109) gave primary importance to the caudal mucus pore. Suter (1892a, p. 270) had proposed the family unit Phenacohelicidae, citing the "caudal gland" as a significant feature, but subsequently (Suter, 1894a, p. 62) agreed with Pilsbry and stated "I do not attach very great importance to the presence or absence of the caudal gland, as we really do not know its true significance; but in the mollusks classed under Flammulina the jaw is always stegognath, the radula is more or less pseudo-zonitoid, and, besides, a mucous tail-gland is always present; whilst in Endodonta and Charopa the jaw is only striated, the radula is much more helicoid, and there is no caudal gland." Suter was using Endodonta in a very broad context, and not in the restricted sense of this study. Early attempts at classifying the Australian and the New Zealand endodontoids were made sequentially in terms of writing, but not in publishing, by Pilsbry (1892a, pp. 54-55), Hedley (1893a, b), Pilsbry (1892b, pp. 68-69), Hedley and Suter (1893, pp. 633-660), and Pilsbry (1893a, pp. 401-402). Pilsbry summarized his views (Pilsbry, 1893-1895, pp. 6-54) in an annotated check list, which included his (Pilsbry, 1893a, pp. 401- 404) placing Laoma and Punctum into a "Group Polyplacognatha" and the remaining into "Group Haplogona" of the Family Endodontidae. The latter name must date from February 2, 1895, the publi- cation date for the introductory pages in that volume. In his monumental survey of New Zealand mollusks, Suter (1913) defined the Phenacohelicidae (p. 621) as with a mucus pore and the Endodontidae (p. 684) as lacking a pore. He divided the latter family into two subfamilies, the Endodontinae with tricuspid lateral teeth and a thin striated jaw, while the Laominae have bicuspid lateral teeth and a jaw of separate plates. Suter (1913, pp. 619-621) also included a peculiar slug-like animal, Otoconcha dimidiata (Pfeiffer, 1853), as a limacid slug, although Cockerell (1893, pp. 188, 205) had placed it in a subfamily, Otoconchinae, without giving any description. Sub- sequently, H. B. Baker (1938a) stated "I am inclined to regard it as constituting an aberrant subfamily of the Endodontidae, but, with almost equal reason, it might be considered as another primitive member of the Arionidae or be erected into a separate family, the Otoconchidae, until intermediate forms are found." Climo (1969a, 1971a) has used Otoconchinae as a subfamily unit and provided much important ana- tomical data on its relatives. Iredale (1913, p. 375; 1915a, p. 479) continued attacking the mucus pore (along with all other anatomical features). Gabriel (1930, pp. 72, 78, 84), in a major paper, proposed the family units Endodontidae, Flammulinidae, and Laomidae for Austalian taxa. Iredale (1937a) adopted this system, without acknowl- edgment, only substituting the name Charopidae for Endodontidae and (Iredale, 1937b, p. 26) adopting Stenopylinae (Thiele, 1931) as a full family unit. Subsequent efforts by Iredale added four undescribed family names, but made no meaningful changes in classification of the Pacific taxa. Climo (1969a, 1970, 1971a, b) has proposed using a single family, Punctidae, with four subfamilies, Charo- pinae, Phenacohelicinae, Punctinae, and Otoconchinae. The two latter are based on the now traditional jaw and radular features (Punctinae) and inevitable consequences of visceral hump reduction (Otocon- chinae). To distinguish between the Charopinae and Phenacohelicinae, Climo relied on the presence (Charo- pinae) or absence (Phenacohelicinae) of an epiphallus. In many taxa with reduced visceral humps, the vas deferens-derived epiphallus will be compacted forward into the penis sheath (Solem, unpublished). Other dissections suggest that the epiphallus in different groups of Australian and New Zealand taxa may be independently derived. In laying to final rest the mucus gland arguments, Climo has performed a notable service, but I do not agree with his criteria for family classification. Only one extralimital paper requires consideration. H. B. Baker (1927, pp. 226-235) reviewed the anatomy and classification of some North and Central Ameri- can endodontoids. His division into the subfamilies Punctinae, Rotadiscinae, and Helicodiscinae was based on changes in pallial cavity configuration and length of the secondary ureter. The pattern of pallial cavity change from Helicodiscus (H. B. Baker, 1927, pi. 18, fig. 42) and Radioconus (pi. 17, fig. 30), to Chan- omphalus (pi. 20, fig. 52), to Radiodiscus (pi. 17, fig. 24), to Punctum (pi. 16, fig. 12), to Rotadiscus (pi. 16, fig. 17) would present a virtually continuous transi- tional series from the pattern found in the Endodon- tidae to that seen in the Charopidae of the Pacific Islands. In addition, Rotadiscus (pi. 16, figs, 13, 19) shows apparently only partial fusion of the prostate and uterus, while the other genera have fused pallial gonoducts. It is premature to try to propose a world- wide classification for this group, since most of the African, South American, Australian, New Caledonian, and Lord Howe Island taxa have not been dissected. The genitalia of the species studied by H. B. Baker (1927) do have typically "charopid" features, so that the pallial configurations sequence does not negate the validity of family level separation. I propose here a three-family classification of the Pacific Basin taxa, into Punctidae, Endodontidae, and Charopidae. The Punctidae have the bicuspid later- omarginal teeth with accessory cusps mentioned above (p. 93) and several differentiating anatomical features that will be discussed elsewhere. The Endodontidae and Charopidae, as represented on the Pacific Islands, differ in a number of major (table LVIII) and minor PHYLOGENY AND CLASSIFICATION 107 (table LIX) anatomical features. There are no known extralimital representatives of the Endodontidae, but the Charopidae have their primary abundance else- where. I include in the Charopidae such taxa as the Phenacohelicinae and Otoconchinae in the sense of Climo (1969a, 1971a), Flammulinidae in the sense of Gabriel (1930) and Iredale (1937a and following), Hedleyoconchidae, and Pseudocharopidae. The ques- tion of subfamily divisions within the Charopidae is deferred until more data are available on Australian and New Caledonian taxa. The relationships of northern hemisphere discids and Neotropical taxa are not discussed at this time. Suter (1913) based family units on the mucus gland; H. B. Baker (1927) based subfamilies on the pallial complex, Thiele (1931) divided the Endodon- tidae into eight subfamilies on shell and radular features, Zilch (1959-1960, pp. 203-230) essentially copied Thiele's classification, except for ranking Otoconchidae as a distinct family, and Climo (1969a) used the presence or absence of an epiphallus for subfamily units. In proposing an increase in rank for units in the classification, as well as altering both the number and composition of these units, I must answer the question as to equivalence with other family units in the Aulacopoda. The characters of major phyletic significance used to separate the Endodontidae from the Charopidae are the absence of the secondary ureter, the complete separation of the prostate and uterus, the very simple structure of the terminal genitalia, and the difference in the ovotestis structure. The other features mentioned in Tables LVIII and LIX are useful, but carry less phyletic weight. The shell structure differences, particularly in the mode of sculpture formation, may have equally significant weight, but need further investigation. The nearest equivalent situation would be the division of the Pacific Island limacaceans into Helica- rionidae and Zonitidae by H. B. Baker (1941, p. 205). His definitions involve divergent patterns of special- ization, such as development of dart apparatus on the female (Helicarionidae) or male (Zonitidae) sides of the terminal genitalia, without listing any equivalent major structural gaps between family units. One of the important changes used here is present within the Helicarionidae. The Microcystinae, the more primitive group that is dominant on the Pacific Islands, has the prostate separated from the uterus, while in the other subfamilies they are united into a "sperm oviduct." But no equivalent of the other major changes exists in the Limacacea. All the limacaceans have a typical sigmu- rethrous pallial complex. There is great specialization of the genitalia including epiphallus formation and (except in the Microcystinae) spermatophore forma- tion with frequent development of accessory genital structure. On the basis of degree of difference, the phyletic gap between the Endodontidae and Charopidae is wider than the gaps between family units of the Limacacea. Much descriptive and some anatomical informa- tion on extralimital Charopidae can be located in faunistic studies. The reports on the molluscan faunas of the Kermadecs (Iredale, 1913, 1915b), Papua (Iredale, 1941c; Solem, 1970a), Lord Howe Island (Iredale, 1944), Norfolk Island (Iredale, 1945), New Caledonia (Solem, 1961), New Zealand (Suter, 1913; Powell, 1957), and the Australian check list (Iredale, 1937a, b, c) provide summaries of the literatures. The incredible nomenclatural nightmare of Iredale (1933) unfortunately cannot be ignored completely, while his subsequent papers on the faunas of New South Wales (Iredale, 1941a, b), and Western Austra- lia (Iredale, 1939) also must be used. The few Philippine Islands (Solem, 1957) and Indonesian (Solem, 1958, 1959b) endodontoids also have been summarized. Connolly (1939) reviewed the South African taxa, and a brief survey of the St. Helena taxa is included in Solem (In press A). Data on Neotropical taxa are very widely scattered. PHYLOGENY WITHIN THE ENDODONTIDAE Perhaps the key problem in phylogenetic analysis today is the question of how to weight characters in determining phylogeny. Opinions vary from the classi- cal pheneticists who stated that every character is of equal weight, to the classical typologists who picked out single characters on which to base decisions. In between are the vast majority of systematists. The present study is more pragmatic than philosophical, although based on the tiered approach to character analysis developed in Solem (In press B). I assume that the major changes involved in progressive evolu- tion require shifts in ecological roles accompanied by morphological alterations. Adaptative radiations with- in such a new zone will involve change at a different level, while the interactions between sympatric species will produce yet a third level of evolutionary change. While the basis of change is genetic, as a practical matter most systematic work must be with morphology, expressed as either a direct or pleiotropic effect of a genetic shift. Biochemical criteria, physi- ological factors, behavior patterns, and molecular data would follow similar patterns. In relation to this study, I consider that the changes from Endodontidae to Charopidae (strictly terrestrial to semi-arboreal, wider tolerance of disturb- ed conditions, changed pallial structures, advanced genital structures) are representative of progressive evolution. No such changes were detected within the Endodontidae, but there are some minor adaptative shifts and numerous instances of sympatric species interactions (see pp. 80-81, tables LII - LIV). Because I have not been able to pinpoint an ancestor group for the Endodontidae, reference to a more primitive outgroup for determination of generalized character 108 SOLEM: ENDODONTOID LAND SNAILS states has not been possible. Instead I have used a short set of pragmatic guidelines. These are based in part on the distributional fact that the Pacific Island endodontoids occur on tiny specks of land that are widely separated from each other. This has the practical effect of making a systematist investigate with great care situations where a species, found on one of the Palau group, for example, has characters that appear very similar to or identical with characters found otherwise only in a Marquesan species. Conti- nental areas have, in many parts of the world, been subject to multiple migrations, invasions, extinctions, and recolonizations because of Pleistocene phenomena. It is intellectually far more satisfying (and comfort- able) to accept disjunct similar species on continental areas as representing distributional relicts of common ancestry than to assume that the Palau and Marque- san species had common ancestry. The basic criteria used in judging change in character states in regard to individual structures or complex patterns of growth are: 1) If formed in exactly the same way they are presumed to have common ancestry; 2) If formed in different ways, although performing the same function or showing the same end growth pattern, they are independently derived; 3) Greater complexity may be suggestive of a derived condition, but if the less complex conditions are non- coherent with each other, while the more complex condition has detailed structural consistency, then secondary simplification is postulated. In regard to distributional factors, I have assumed that: 4) Widely distributed character complexes that have structural consistency probably are ancestral to sporadically distributed different states of these complexes that lack structural consistency. 5) Character states of limited geographic occurrence should be analyzed in terms of development from or into states of wide geographic distribution. 6) Character states must be interpreted also in reference to conditions existing among sympatric or probably sympatric taxa. 7) Character states occurring in only one geographic area may be either generalized or derived in comparison with widely distributed states, with interpretation resting on correlated changes with other characters that can be interpreted more objectively. Examples of the ways in which these criteria have been applied during this study are: 1) and 2) The apertural barriers in the Endodon- tidae have the same type of microdenticulations on their upper surface and therefore the barriers are assumed to be of common origin, while the barriers in the Charopidae show different types of structure and superior microdenticulations, strongly suggest- ing multiple origins (Solem, 1973b, p. 305). The brood chamber growth pattern in the Endodontidae occurs in several different geographic areas. In each situation the method of secondarily narrowing the umbilicus is different, suggesting multiple origin of the growth pattern (pp. 27-30). 3) Reduction of the apertural barriers results in a very simple ridgelike structure, particularly when compared with the detailed structures found on the larger barriers. As shown above (pp. 57,62), the patterns of reduced barriers are much more varied than are the patterns of fully developed barriers. Reduced shell sculpture (pp. 47-50) correlates with increased shell size and the patterns of reductions have greater variability than do the basic complex sculpture. 4) The pattern of the penis with two low pilasters, the vas deferens entering below the apex, and penial retractor muscle inserting on the penis apex is widely distributed, while the additions of epiphallic tissue to the penis apex and changes in the pilaster patterns occur sporadically. 5) and 7) The presence of a glandular zone inside the penes of Marquesan Endodontidae has no counter- part elsewhere in the family, and, if eliminated, the penis structure would still be specialized in terms of the family pattern. Hence this is interpreted as an additive, specialized structure. 6) The variations in penis size and pilaster patterns (tables LII, LIII, LIV) are largest when sympatry of congeners is involved. Hence aberrant structural patterns in the penis complex are viewed first as suggesting "species recognition" interactions be- tween populations. Comparisons must be made with sympatric or at least same-island taxa before predicting whether the variation represents a general adaptational trend or essentially local character displacement to aid species recognition. All of the above guidelines are based on the attempt to understand the ontogenetic development of structures and to place them within the framework of species-level interactions. This approach is more applicable to mollusks than to arthropods or verte- brates, since the ontogenetic pattern of shell growth is available in each adult specimen, while obviously lacking in the adult arthropod or vertebrate. Both the key to and difficulty of this approach involve the necessity to interpret not just the final structure, but to analyze its components and ontogeny as an aid toward deciding its significance in phylogenetic analysis. PHYLOGENY AND CLASSIFICATION 109 TABLE LX. - SHELL PARAMETERS OF THE ENDODONTIDAE Minimum 1st Quartile Median 3rd Quartile Maximum Height 0.92 1.58 1.98 2.57 7-26 Diameter 1.68 3.01 3.77 It. 8 5 12.26 H/D ratio 0.3^ 0.1+66 0.531 0.589 0.789 Whorls 3-5/8 5-1/8 5-1/2+ 6-3/8- 8+ D/U ratio 1 1.68 3. 1 1 * 3. 9^ 5.6l closed Ribs 2 19 63.6 80.0 IQh.k 250 Ribs /mm. 1.1*1 5-0 7.6 11.1 1+0.1+ Excluding brood chamber taxa Excluding those without countable ribbing on body whorl Portrait of a generalized endodontid Although the most basic trends in the Endodon- tidae are toward increased size accompanied by structural alterations, there is evidence that at least one genus, Rhysoconcha, is secondarily dwarfed (pp. 255-256). It should not be assumed automatically that the smallest species in size represent the most generalized taxa. A reduction in adult whorl count also can effectively produce smaller adult size without requiring major structural alterations. Table LX summarizes the distribution of several shell parameters in the Endodontidae. Allowing for the tendency toward larger size that often is the pattern in most taxa, and utilizing data from the discussion given above on variation in structures, the "generalized endodontid" structure can be described rather simply. The shell would be about 3.0-3.5 mm. in diameter, with 5 1 A - 5!/2 whorls, the height being slightly less than half the shell diameter, and the widely open umbilicus would be contained about 3.5 times in the diameter. There would be a prominent sculpture, numbering perhaps 65-90 ribs on the body whorl, spaced six to eight per millimeter of shell periphery, and with four to eight microradials between each pair of major ribs. The apex and spire would be slightly to moderately elevated, reaching up to half the body whorl width in terms of actual protrusion. The body whorl itself would be rounded or laterally flattened. Inside the aperture there would be 2 to 3 parietal barriers extending three-sixteenths to one-quarter whorl posteriorly, 1 columellar barrier, and 4 palatal barriers at or near the lip edge that extended one-eighth to three-sixteenths whorl posteriorly. All the barriers would be slightly to moderately widened above on the posterior half to two-thirds, and capped on the expanded portions with triangular microdenticulations that point toward the outside of the aperture. The anterior portion of each barrier would gradually descend to a sharper truncation in many palatals and an anterior threadlike portion in the lower parietals. The net effect of these barriers would be to grossly narrow the apertural opening, except for a slightly widened zone in the upper palatal area to permit effective withdrawal of the head and foot. In the anatomy, only the few variable features need to be outlined, since the structural plan is relatively uniform. The pallial cavity would extend about three-quarters whorl, with the kidney weakly bilobed, and probably there would be no mantle gland extension onto the pallial roof. The genitalia would have the penial retractor muscle inserting directly on the head of the penis and originating from the diaphragm. Inside the penis would be two low and rounded, longitudinal pilasters, with the vas deferens opening just below their point of apical union. The spermathecal insertion is uncertain, and other features of the genital system seem to vary more in size correlated features than anything else. The radula would have about 100 rows of teeth, with a tricuspid central, five or six bicuspid laterals, and perhaps 10-12 marginals with split cusps. The jaw would be composed of separate, elongated plates held together by a thin membrane. Converting the above description into an ancestor of the present endodontids probably would involve only a reduction in shell diameter and whorl count, with correlated changes in ribbing, D/U ratio, and pallial length. Identifiable major trends Smaller shell size can be reached by the Rhyso- concha strategy of secondary size reduction without major reduction in whorl count, or by the pattern that may exist in the smallest Minidonta, Cookeconcha, and Mautodontha where greatly reduced whorl counts are common. By far the most prevalent trend is for increase in shell size, most often by simple continued addition of more whorls to the shell. Associated with this increased size are a tendency for loss of shell sculpture, fusion of the jaw plates, often a more elevated spire as the decoiling growth continues, sometimes lengthening and size reduction of the 110 SOLEM: ENDODONTOID LAND SNAILS I \ Orangia [ Rhysoconcha [/ *r ^*"^ *r I Kondocon FIG. 57. Phyletic diagram of the Endodontidae showing hypothesized origins of extant taxa. apertural barriers, and various minor lengthening trends as outlined in discussions of the characters above. Such simple continued incremental growth in whorl numbers also will tend toward a very widely open umbilicus and proportionately higher shell that may reduce the ability of the animal to crawl into narrow crevices. The widely open umbilicus partic- ularly may have led to a functional problem with egg retention. In some Nesodiscus (p. 345) eggs deposited in the widely open umbilicus are covered by a solid sheet of mucus to form an encapsulated situation, but in at least five groups (p. 28) there has been secondary narrowing of the umbilicus to form a "brood chamber" in which the eggs are deposited. Inevitably, widening of the umbilicus had to precede the secondary narrowing (fig. 189), and it is the differences in the way that the narrowing is achieved that indicate separate origin of this growth pattern. These then are two basic trends in shell variation size increase through whorl accretion leading to first a very widely open shell umbilicus and a tendency toward secondary narrowing of the umbilical cavity to form a brood chamber. Anatomical variations are, in general, correlated with minor features of shell variation. Only a few seem to be independent of the trends in shell structure. Ignoring the "species recognition" changes in genital structure, the one really striking alteration is in the pattern of the penis-vas deferens-epiphallus relation- ship. This is geographically limited. Many Rapan and most Marquesan taxa show an added zone of glandular tissue to the penis apex (fig. 46), while the Palau Island Aaadonta, Fijian Priceconcha, Zyzzyx- donta, and Thaumatodon have a quite different entrance of the vas deferens and altered attachment of the penial retractor, that, in effect, forms a penial epiphallic section. Other local changes in anatomical structure that can be called trends are the addition of a pustulate zone within the penis of Marquesan taxa, and the tendency for increase in radular tooth numbers for some Hawaiian and Marquesan taxa. While many features of the shell sculpture and apertural barriers were shown above to be partly size correlated, others vary more within a phyletic unit on a geographic basis. The obvious change in barrier microdenticulation from a continuous surface in most genera to the "beaded" structures seen in the Aaadonta-Thaumatodon group correlates with the penial epiphallus grouping, and represents a major change in structure. As an example of retaining a marked pattern of variation through major shell size and shape shifts, the very characteristic apertural barriers of Anceyodonta also are seen in the Man- garevan Minidonta and Gambiodonta. All the Man- garevan taxa also have the tendency to develop microdenticulated trace barriers (figs. 71c; 89d, f; 187). Phylogenetic conclusions The Pacific Island Endodontidae are characterized by a repetitive set of conchological specializations that have produced frequent convergences in appearance, plus a few anatomical trends that do not correlate with the basic conchological trends. If more taxa had been available for dissection, particularly from areas such as Mangareva and the Society Islands, the number of identified anatomical trends undoubtedly would have increased. My ideas concerning the phylogeny of the Endodontidae are summarized in Figures 57 and 58. The first is a typical phyletic tree Q .'.............. UILU U| 111 112 SOLEM: ENDODONTOID LAND SNAILS diagram, while the second introduces additional data concerning degree of relationships and patterns of specialization. Detailed arguments concerning inter- generic relationships are developed in the systematic review and are not repeated here. The first conclusion, shown in Figure 57, is that Cookeconcha and Minidonta, although now showing divergent patterns of structure, shared a direct common ancestor and formed the stem groups for subsequent evolution. That they are truly primitive in all features is unlikely, since the unusual umbilical decoiling pattern of many Minidonta (figs. 62c, f; 63b; 69c, f) and the bifid parietal of many Cookeconcha represent highly atypical situations within the context of the family. Nevertheless, their patterns of dis- tribution and most structural features place them perhaps nearest to the potential ancestral group of any extant taxa. The Hawaiian radiation into Nesophila and Endodonta is derived from Cookeconcha. Un- doubtedly more generic level taxa will be delineated when the Hawaiian fauna has been reviewed in detail. Mautodontha, Australdonta, and Anceyodonta represent separate specializations from the Minidonta complex, each subsequently giving rise to other taxa. The Society, Cook, and Tuamotu island Libera, Kleokyphus, Nesodiscus, and Pseudolibera represent local developments from subgenera of Mautodontha. The Marquesan Taipidon also is derivable from very generalized Mautodontha, but shows a variety of specializations, one of which led to the endemic Planudonta. All available data suggest that the Rapa Island radiation is monophyletic, with Opanara repre- senting the generalized condition from which Rhyso- concha, Ruatara, Orangia, and Kondoconcha were independently derived. Opanara more probably was derived from a Mautodontha-type ancestor, but could have evolved from a Minidonta-\eve\ ancestor of large size. On Mangareva in the Gambler Islands, Rikitea and Gambiodonta represent local derivates either directly from the Minidonta stock or from Anceyo- donta. As outlined in the zoogeography section (pp. 488- 492), the Thaumatodon-Aaadonta complex has a very different pattern of distribution from that shown by the other taxa. It also shows significant changes in structure. I regard it as being a more recent element in the Pacific Island fauna, one that evolved from endodontids formerly in the Indonesia-New Guinea- Australia axis that are now extinct, their place having been taken by the Charopidae. While Thaumatodon and its derivatives share a distant common ancestry with the other species groups, this complex probably is not derived from any of the extant Pacific Island taxa, but rather from extralimital groups. Although the structures found in the Thaumatodon-Aaadonta complex can be derived from those found in the Mautodontha-level taxa, the discordance in dis- tribution type is so large and the morphological gap so abrupt that the hypothesis of an independent origin followed by secondary colonization seems far more probable to me. A different approach is taken in Figure 58 which indicates four levels of conchological specialization that are in great part size correlated, presents additional data about interrelationships of species groups, summarizes size range within genera, and also includes data on probable directions of evolution within the group. One error resulting from lay-out problems and one omission need to be mentioned. Libera probably is derived from a Garrettoconcha-type ancestor rather than Mautodontha, s. s., and Price- concha should have been shown as coming from the Thaumatodon stem as a "Nesodiscus level" taxon right under the "Zyzzyxdonta" label. Inclusion of the recently described Priceconcha would have required redoing the entire chart. A brief outline of the characteristics for each of the four specialization levels follows. In increasing order of specialization, they are the Minidonta, Mautodontha, Nesodiscus, and brood-chamber levels. Three genera are included on the Minidonta level. These include the genus Rhysoconcha by secondary derivation, the most generalized species group of Cookeconcha (excluding C. nudus), and Minidonta itself. As summarized in Table LXI, these species show an average H/D ratio, but fall into the lower quartile in both diameter and whorl count (cf. table LX). The umbilicus is slightly narrower than average, relating to both the low whorl count and the peculiar pattern of umbilical decoiling seen in some Minidonta. All of these taxa retain a prominent shell sculpture, and the vast majority (16 of 21) have the aperture moderately to strongly constricted by the barriers. Minidonta and the most generalized Cookeconcha closely approach each other in overall structure, but there are numer- ous characters in which they contrast, as discussed under Cookeconcha subpacificus on pp. 211-212. The other Hawaiian genera are derivable from the general- ized Cookeconcha, while the species groups within Minidonta serve as effective stem groups for many other taxa. Minidonta grades almost imperceptibly into the Mangarevan Anceyodonta (pp. 179-181), while Australdonta can be derived from the M. anatonuana complex. The Rapan radiation could be descended from either Minidonta or, more probably, the Mau- todontha complex, as indicated by the dotted lines. Mautodontha is, in itself, a stem group for a wide variety of taxa. Estimating the exact relationships between the more specialized Mautodontha-level taxa and the species groups clustered as Minidonta is hampered by the virtual lack of any anatomical data for species in either Minidonta or Mautodontha. A simpler classification would result if the geographical species groups of Minidonta were associated with their geographical derivatives, to form linear genera. I have not done so, since the morphologic gap between the derived genera and the Minidonta-species groups that are logical ancestors to them usually is greater than PHYLOGENY AND CLASSIFICATION 113 TABLE LXI. - SHELL PARAMETERS FOR LEVELS OF ORGANIZATION Number of taxa Minidonta 21 Mautodontha 111* 3.75 0.098 Nesodiscus IT Brood chamber 29 Diameter 2.25 0.095 3.75 0.098 5.90 0.538 5-39 0.277 (1.68-3.26) (1.87-8.99) (3.75-11.29) (U. 23-12. 26) H/D ratio 0.5^5 0.010 0.528 0.009 0.1+1+9 0.016 0.597 0.012 (0.1+1+5-0.625) (0.3^-0.789) (0.31+6-0.560) (0. 1+80-0. 702) Whorls 1+.83 0.12 5- 1 +9 0.06 (3-5/8 - 5-1/2+) (U-8) D/U ratio 1+.87 0.39 (2.66-10.1) the gaps between the species groups clustered within Minidonta. The Mautodontha level of organization contains the bulk of the species and genera. They show the median pattern in size, H/D ratio, and whorl count (tables LX, LXI). Since species with closed umbilici fall into this grouping, the mean D/U ratio was not calculated. Currently, the stem genus, Mautodontha, is virtually geographically isolated from Minidonta, except for the joint occurrence on Raivavae in the Austral Islands caused by the inclusion of Mau- todontha ceuthma in that genus. This means that I consider Mautodontha to be potentially a grade, rather than a clade. With the lack of anatomical data and limited material available, I have included M. ceuthma in Mautodontha rather than with the derivative Australdonta. Discussions of the derivation patterns for Kleokyphus, Nesodiscus, Pseudolibera, Libera, Taipidon, and Australdonta are discussed under the respective genera. Several groups show general trends toward, or include species that actually have reached, the more specialized levels. For example, Taipidon semimarsupialis has a definite brood cham- ber, although no other member of the genus comes close to attaining this level of specialization. In contrast, while only Endodonta marsupialis has secondary umbilical narrowing, most of the other taxa have a U-shaped, deep umbilicus and are perfectly "pre-adapted" to a narrowing trend. Kleokyphus from Makatea and Kondoconcha from Rapa also are close to this specialization pattern. Yet another taxon, Thaumatodon euaensis, also shows umbilical narrow- ing, but never had enough umbilical expansion to justify calling this a brood chamber development. Australdonta pharcata, A. ectopia, Opanara m. megomphala, and O. m. tepiahuensis are Mau- todontha-level taxa that show or approach the Nesodiscus pattern of specialization. The Nesodiscus-\eve\ genera, Nesophila, Neso- discus, Planudonta, and Priceconcha are quite strongly characterized. There is a gross reduction in 6.1+0 0.19 6.78 0.11 (U-7/8 - 7-3/8) (5-3/8 - 8+) 2.21+ 0.10 (1.68-3.11+) sculptural prominence (less in Planudonta), usually great-to-complete reduction in the apertural barriers, a sharp increase in diameter (table LXI) accompanied by an extremely widely open umbilicus, a whorl count mostly in the upper quartile, but also a lower quartile H/D ratio. These species have increased their whorl count in regular fashion, but with umbilical widening not followed by a change in growth vectors to produce either a brood chamber or a very high spire. The trends to sculpture reduction and loss or reduction of the apertural barriers are quite consistent and contrast with the pattern in the brood-chamber taxa, where sculpture reduction is much less frequent and the size reduction in the parietal barriers far less accentuated. The method of sculpture reduction also differs. In the Nesodiscus level this occurs first by multiplication of rib numbers and crowding, followed by their loss (except in Planudonta). In brood-chamber taxa rib loss occurs by gradual size reduction in the major ribs, rather than multiplication, then loss. In calculating the averages for the Nesodiscus level, I omitted the Australdonta and Opanara species listed above, since they agree only with part of the character complex. One additional genus requires comment. The very poorly known Mangarevan taxon Rikitea has the shape and growth pattern of the Nesodiscus special- ization, but differs quite obviously in retaining a very large parietal barrier and prominent radial ribbing. The two Australdonta (fig. 137) come much closer to reaching "Nesodiscus status." The brood-chamber taxa, Libera, Gambiodonta, Pseudolibera, Endodonta marsupialis, and Taipidon semimarsupialis, show a continued increase in whorl count, an H/D ratio in the upper quartile, and a lower diameter than the Nesodiscus-level taxa. These are functional requirements of this level, since secondary narrowing of the brood chamber necessitates shifting growth vectors to increase the shell height and lessen the diameter. Their retention of strong sculpture and prominent apertural barriers, in contrast to the Nesodiscus series, suggests that these represent paral- lel rather than sequential stages. While umbilical 114 SOLEM: ENDODONTOID LAND SNAILS widening in the Nesodiscus pattern is a mandatory prelude to secondary narrowing and brood chamber formation, the derivation of Nesodiscus from a species very similar to Mautodontha boraborensis is discussed below (p. 345) and the derivation of Libera from the subgenus Garrettoconcha (p. 165) hints at the contrasting patterns. Whether the sculpture reduction, barrier reduction and continued umbilical widening of the Nesodiscus are genetically linked in a formal sense or became linked in a channeled development pattern is unknown. It is probable, however, that no Neso- discus-level taxon would alter its pattern to shift into the U-shaped and then narrowed umbilicus seen in brood-chamber taxa. In contrast, the tendency of many Mautodontha-leve\ taxa to form U-shaped umbilici while retaining heavy sculpture and promi- nent barriers suggests that there is more than chance to these divergent patterns. Most detailed discussion of phylogeny in the systematic review is based on geographic lineages, since the patterns in variation of the apertural barriers and anatomy that are not size correlated clearly link together the taxa from each island group. For example, in the Hawaiian genera there are the common patterns of bifidity in the barriers, shift in radular tooth shape, and standard penis structure; the Mangarevan taxa have a characteristic tooth structure that encompasses all material except the peculiar Rikitea; the Marque- san genera have the altered penis structure; and the Thaumatodon-Aaadonta complex has not only the altered penis but also the very striking change in barrier expanded surfaces. The convergences in the shell size, shape, and sculpture that mostly correlate with size factors are extensive enough that they swamp the few factors used to establish phyletic affinity if all are tossed into a phenetic program. During the middle portion of this study, I was able to have data on the shell variables put through both the "minimum steps" and Sharrock-Felsenstein "combinatorial" programs for computing phylogenies. Forty-seven meristic or structural features were coded and directional changes from the postulated "general- ized" condition indicated. The computer programs available for use at that time required data on every species for each character used, which effectively eliminated using any anatomical data, and could handle only 25 taxa at a time. The characters used were the shell features discussed above (pp. 19-72). The only difference was that my analysis of the ways in which the same state could be achieved independently had not been carried nearly as far, nor were character correlations as fully understood. As would be expected, the programs separated out highly differentiated taxa, such as distinguishing between Thaumatodon, Aaa- 5 10 11 12 9 13 14 15 16 17 ancestor ancestor Fu;. 59. Computer generated phylogeny, "combinatorial" method, of Thaumatodon, Zyzzyxdonta, and Aaadonta. Derivation from left to right in sequence. Extant species are: 1) Thaumatodon multilamellata; 2) T. decemplicata; 3) T. laddi; 4) T. subdaedalea; 5) T. corrugata; 6) T. hystricelloides; 7) T. vavauensis; 8) T. euaensis; 9) Zyzzyxdonta alata; 10) Aaadonta pelewana; 11) A. f. fuscozonata; 12) A. f. depressa; 13) A. kinlochi; 14) A. irregularis; 15) A. c. komahanensis; 16) A. c. babelthuapi; 17) A. c. constricta; 18) A. angaurana. Small numbers are Thaumatodon; large numbers Aaadonta; bold face Zyzzyxdonta. a and 6 represent different computer runs. me HI? OJ CJ o OJ ^3 LT . O QO Hill S 9* Si (\ '." i'^JI* .00 S JPt: S ^ a C c-i 5 -H N .Q H 115 116 SOLEM: ENDODONTOID LAND SNAILS donta, and Zyzzyxdonta (fig. 59), but when presented with somewhat similar taxa from different islands, such as Orangia, Opanara, and Rhysoconcha from Rapa, plus Taipidon from the Marquesas (fig. 60), or Mlnidonta, Mautodontha, and Cookeconcha subpa- cificus (fig. 61), the results were less satisfactory. The differentiated taxa run (fig. 59) had Zyzzyx- donta (species 9) from Fiji associated with Aaadonta kinlochi (species 13) from Palau. Both species are low- spired, carinated, and show quite different proportions from typical members of either genus. Different computer runs (fig. 59a, b) give different results. A. angaurana (species 18) could be associated with either A. c. constricta (species 17 in fig. 59a) or A. irregularis (species 14 in fig. 59b). Both runs mixed up the Thaumatodon geographically, grouping Fiji (3) and Ellice (2), Samoa (6) and Tonga (7) in Figure 59a, and partly in Figure 59b. Both runs segregated the races of Aaadonta constricta and recognized A. pelewana and A. fuscozonata as a monophyletic assemblage. The Taipidon and generalized Rapa Island taxa (fig. 60a, b) included only situations where the genera are very well differentiated by anatomical criteria. The shells show a wide variety of convergences. In the absence of anatomical data it would be unreasonable to expect that the computer would avoid confusing convergence with phyletic affinity. It did not. Linking Opanara megomphala (species 12) with Taipidon centadentata (species 8 in fig. 60b) is quite logical on overall appearance, provided no anatomical data is available. The extensive mixing of geographic and generic groups in the Minidonta-Mautodontha run (fig. 61) requires no commentary. What is intriguing in this phylogeny is the placement of Cookeconcha subpa- cificus (species 1) in basal position, the grouping of Kleokyphus (species 24, 25) as highly specialized taxa, and the two peculiar Aitutaki species (13, 23) as highly derived taxa sharing many similarities. Shell data alone, particularly when convergent variations are not carefully screened out, are quite inadequate to permit obtaining "good phylogenies" in the Endodontidae by computer analysis. The fault lies not with the computer, but with the subtle and repetitive nature of the variations. The island dis- tributions were an incalculably great aid to this study. If a continental pattern had been involved, I doubt very much that I would have been able to recognize patterns so clearly. The results of both the conventional and computer studies do emphasize the need to understand the functional significance of character variation and to weight the variations accordingly. The computer is an invaluable tool in sorting taxa, suggesting possible relationships, empha- sizing convergences, and forcing one to examine and analyze far more data than previously. But it is no substitute for the more conventional approaches used here. The above discussion completes the tentative review of endodontid phylogeny. As a final and totally subjective comment, there are geographical "styles" of variation. While the Mangarevan and Society Islands differentiations into more specialized levels are "ma- ture" in character, the Marquesan experiments into brood chamber and Nesodiscus level seem tentative and "juvenile." The end results were achieved in Taipidon semimarsupialis and Planudonta, respective- ly, but in far less polished ways than is shown by the other groups. The Hawaiian taxa show almost an exuberant pattern of minor experimentations, which is equalled by Thaumatodon and its derivatives in the Lau Archipelago of Fiji. In contrast, the divergence of Aaadonta in Palau shows only minor variations on a theme, while on Rapa there has been variation more in anatomy than shell structure. Quite possibly this is a function of island age and time of colonization, but such zoogeographic topics are deferred. FOSSIL ENDODONTOID LAND SNAILS In favorable circumstances, fossils can yield data that are crucial to interpreting phylogeny and estimat- ing rates of evolution. Unfortunately, the fossil endodontoid land snails (Ladd, 1958, 1968; Ladd et al., 1967, 1970) add details rather than providing major input. At present there are six taxa known, two charopids, "Ptychodon" eniwetokensis Ladd (1958) and "P." davidi Ladd (1968), two endodontids that are reviewed below (Minidonta inexpectans, p. 132, and Cookeconcha subpacificus, p. 212), and two unde- scribed species from the core drillings on Midway. These range in age from Lower Miocene to Late Pleistocene or Recent. They are discussed in order of decreasing age. The oldest species, Cookeconcha sub- pacificus from the Lower Miocene of Bikini Atoll at 1,807-1,818 ft. is based on a fragmentary specimen that agrees most closely with the Cookeconcha henshawi group (p. 213). None of the preserved features on this shell are inconsistent with extending Cookeconcha back to the Lower Miocene and from the present Hawaiian range to include the Marshall Islands. "Ptychodon" eniwetokensis, from the Upper Miocene of Eniwetok at 820-831 ft., also is based on a fragmentary example, but the available features place it in a relatively advanced genus (undescribed) of the Charopidae. The Eniwetok fossil agrees more with extant congeneric taxa from Niue and Vaitupu in the Ellice Islands than with the species from Fiji and Tonga. The latter are more specialized in structure. The three more generalized taxa thus form a rough fringing pattern of distribution within the genus. The range extension from Vaitupu to Eniwetok is not very significant in terms of geography, although placing the fringe distribution in an Upper Miocene context has some importance. The Pliocene to Pleistocene Min- idonta inexpectans from Bikini at the 447-453 ft. level is very close to the recent Samoan Minidonta manuaensis (pp. 130-132). Together with Minidonta IDi 12 ||LO \ o v> c_> cr CO III