Phylogenetic relationships of Icaronycteris, Archaeonycteris, Hassianycteris, and Palaeochiropteryx to extant bat lineages, with comments on the evolution of echolocation and foraging strategies in Microchiroptera.

2 Icaronycteris index (holotype; PU 18150) from the Green River Formation, Wyoming. CONTENTS Abstract ....................................................................... 4 Introduction .................................................................... 5 Relationships and Classification of Eocene Bats: A Historical Overview ........... 12 Relationships Among Extant Lineages of Bats .................................. 31 Goals of the Current Study ................................................... 40 Materials and Methods ......................................................... 40 Taxonomic Sampling and Monophyly .......................................... 40 Outgroups .................................................................. 42 The Data Set ................................................................ 46 Characters Examined in Fossil Bats ............................................ 48 Skull and Dentition ........................................................ 48 Anterior Axial Skeleton .................................................... 64 Pectoral Girdle ............................................................ 70 Forelimb ................................................................. 77 Posterior Axial Skeleton and Pelvis .......................................... 81 Hindlimb ................................................................. 84 Completeness ............................................................... 85 Methods of Phylogenetic Analysis ............................................. 86 Analysis of Character Transformations ......................................... 87 Phylogenetic Results ........................................................... 88 Results of Analyses .......................................................... 88 Analysis 1: All Characters, Fossil Taxa Excluded .............................. 88 Analysis 2: All Characters, All Taxa ......................................... 89 Analysis 3: All Taxa, Characters Limited to Those Scored in Fossil Forms ....... 92 Summary ................................................................... 94 Character Evolution in Early Chiropterans ........................................ 94 Character Transformations Associated with Basal Nodes .......................... 94 Features Diagnosing the Microchiropteran Crown Group ......................... 99 Character Transformations at Basal Nodes: A Functional Perspective ............. 101 Evolution of Echolocation and Foraging Strategies ................................ 107 Timing of the Origin of Flight and Echolocation ............................... 108 Vision and the Evolution Of Echolocation ..................................... 111 Foraging Ecology of Eocene Bats ............................................ 119 Evolution of Foraging Strategies: A Phylogenetic Perspective .................... 129 Classification of Eocene Bats ................................................... 133 Conclusions and Summary ..................................................... 139 Acknowledgments ............................................................ 143 References ................................................................... 143 Appendix 1: Specimens Examined .............................................. 170 Appendix 2; Character Descriptions ............................................. 172 Appendix 3: Data Matrix ...................................................... 180 4 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY ABSTRACT The Eocene fossil record of bats (Chiroptera) includes four genera known from relatively complete skeletons: Icaronycteris,Archaeonycteris, Hassianycteris, and Palaeochiropteryx. Phylogenetic relationships of these taxa to each other and to extant lineages of bats were investigated in a parsimony analysis of 195 morphological characters, 12 rDNA restriction site characters, and one character based on the number of R-1 tandem repeats in the mtDNA d-loop region. Results indicate that Icaronycteris,Archaeonycteris,Hassianycteris, and Palaeochiropteryx represent a series of consecutive sister-taxa to extant microchiropteran bats. This conclusion stands in contrast to previous suggestions that these fossil forms represent either a primitive grade ancestral to both Megachiroptera and Microchiroptera (e.g., Eochiroptera) or a separate clade within Microchiroptera (e.g., Palaeochiropterygoidea). A new higherlevel classification is proposed to better reflect hypothesized relationships among Eocene fossil bats and extant taxa. Critical features of this classification include restriction of Microchiroptera to the smallest clade that includes all extant bats that use sophisticated echolocation (Emballonuridae 1 Yinochiroptera 1 Yangochiroptera), and formal recognition of two more inclusive clades that encompass Microchiroptera plus the four fossil genera. Comparisons of results of separate phylogenetic analyses including and subsequently excluding the fossil taxa indicate that inclusion of the fossils changes the results in two ways: (1) altering perceived relationships among extant forms at a few poorly supported nodes; and (2) reducing perceived support for some nodes near the base of the tree. Inclusion of the fossils affects some character polarities (hence slightly changing tree topology), and also changes the levels at which transformations appear to apply (hence altering perceived support for some clades). Results of an additional phylogenetic analysis in which soft-tissue and molecular characters were excluded from consideration indicate that these characters are critical for determination of relationships among extant lineages. Our phylogeny provides a basis for evaluating previous hypotheses on the evolution of flight, echolocation, and foraging strategies. We propose that flight evolved before echolocation, and that the first bats used vision for orientation in their arboreal/aerial environment. The evolution of flight was followed by the origin of low-duty-cycle laryngeal echolocation in early members of the microchiropteran lineage. This system was most likely simple at first, permitting orientation and obstacle detection but not detection or tracking of airborne prey. Owing to the mechanical coupling of ventilation and flight, the energy costs of echolocation to flying bats were relatively low. In contrast, the benefits of aerial insectivory were substantial, and a more sophisticated low-duty-cycle echolocation system capable of detecting, tracking, and assessing airborne prey subsequently evolved rapidly. The need for an increasingly derived auditory system, together with limits on body size imposed by the mechanics of flight, echolocation, and prey capture, may have resulted in reduction and simplification of the visual system as echolocation became increasingly important. Our analysis confirms previous suggestions that Icaronycteris,Archaeonycteris,Hassianycteris, and Palaeochiropteryx used echolocation. Foraging strategies of these forms were reconstructed based on postcranial osteology and wing form, cochlear size, and stomach contents. In the context of our phylogeny, we suggest that foraging behavior in the microchiropteran lineage evolved in a series of steps: (1) gleaning food objects during short flights from a perch using vision for orientation and obstacle detection; prey detection by passive means, including vision and/or listening for prey-generated sounds (no known examples in fossil record); (2) gleaning stationary prey from a perch using echolocation and vision for orientation and obstacle detection; prey detection by passive means (Icaronycteris,Archaeonycteris); (3) perch hunting for both stationary and flying prey using echolocation and vision for orientation and obstacle detection; prey detection and tracking using echolocation for flying prey and passive means for stationary prey (no known example, although Icaronycteris and/or Archaeonycteris may have done this at times); (4) combined perch hunting and continuous aerial hawking using echolocation and vision for orientation and obstacle detection; prey detection and tracking using echolocation for flying prey and passive means for stationary prey; calcar-supported uropatagium used for prey capture (common ancestor of Hassianycteris and Palaeochiropteryx; retained in Palaeochiropteryx); (5) exclusive reliance on continuous aerial hawking using echolocation and vision for orientation and obstacle detection; prey detection and tracking using echolocation (Hassianycteris; common ancestor of Microchiroptera). The transition to using echolocation to detect and track prey would have been difficult in cluttered envionments owing to interference produced by multiple returning echoes. We therefore propose that this transition occurred in bats that foraged in forest gaps and along the edges of lakes and rivers 1998 5SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS in situations where potential perch sites were adjacent to relatively clutter-free open spaces. Aerial hawking using echolocation to detect, track, and evalute prey was apparently the primitive foraging strategy for Microchiroptera. This implies that gleaning, passive prey detection, and perch hunting among extant microchiropterans are secondarily derived specializations rather than retentions of primitive habits. Each of these habits has apparently evolved multiple times. The evolution of continuous aerial hawking may have been the ‘‘key innovation’’ responsible for the burst of diversification in microchiropteran bats that occurred during the Eocene. Fossils referable to six major extant lineages are known from Middle–Late Eocene deposits, and reconstruction of ghost lineages leads to the conclusion that at least seven more extant lineages were minimally present by the end of the Eocene. Only extensive phylogenetic analysis, based on as many suites of characters as possible, and carried out in conjunction with adaptational and aerodynamic studies, can form the basis for reconstruction of evolutionary change. Padian (1987: 19) The behavior of fossil animals and the evolution of flight will probably always be a subject of contention, and although we may never know for certain if we have found the right answers, we can always distinguish the possible from the impossible, the probable from the improbable. Norberg (1990: 268) INTRODUCTION Bats first appear in the fossil record in Early Eocene deposits of North America, Europe, Africa, and Australia (table 1). The diversity of Eocene bats is remarkable—24 genera are currently recognized, and new species are described almost every year. Of the eight bat genera that make their first appearance in the Early Eocene, almost half are known from spectacular, nearly complete skeletons: Icaronycteris,Archaeonycteris, Hassianycteris, and Palaeochiropteryx. Much of what is known (or hypothesized) about the early evolution of bats is based on studies of these taxa (e.g., Jepsen, 1966, 1970; Russell and Sige´, 1970; Richter and Storch, 1980; Novacek 1985a, 1987; Habersetzer and Storch, 1987, 1988, 1989; Habersetzer et al., 1989, 1992, 1994; Norberg, 1989; Storch, 1989). Icaronycteris was described by Jepsen (1966) on the basis of a beautifully preserved skeleton from the Early Eocene Green River Formation of Fossil Basin, Wyoming (see Frontispiece). A small bat with a long tail, Icaronycteris had a wingspan of approximately 30 cm and probably weighed 10–16 g (Habersetzer and Storch, 1987; Norberg, 1989). The holotype was collected from beds that are Middle Wasatchian in age, approximately 53 Ma (Woodburne, 1987; Woodburne and Swisher, 1995). At least three additional skeletons of Icaronycteris were subsequently discovered in the same deposits (Novacek, 1985a, 1987; Habersetzer and Storch, 1987; see appendix 1). Isolated teeth from Clarkforkian deposits in North America were referred to cf. Icaronycteris sp. by Gingerich (1987), thus potentially extending the range of the genus to the Late Paleocene (fig. 1). Other fragmentary material suggests that Icaronycteris may have persisted in North America until the Gardenerbuttian (McKenna and Bell, 1997), an early Bridgerian interval that ended approximately 50 Ma (Woodburne, 1987; Woodburne and Swisher, 1995). Only one species, Icaronycteris index Jepsen, 1966, is currently recognized in North America. However, the geographic range of Icaronycteris may have extended beyond North America to Europe. Russell et al. (1973) recognized a new species (Icaronycteris?menui) based on a large collection of jaw fragments and isolated teeth from Early Eocene (fig. 1; MP 1 8–9) deposits in 1 Mammal Paleogene (MP) reference levels are biostratigraphic intervals used to subdivide the Paleogene of Europe. These units, which are of unspecified duration, are based on terrestrial mammalian faunas. The MP system is based on the assumption that the fossil record 6 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY France. However, the affinities of this material remain questionable because of the fragmentary nature of available specimens and lack of any diagnostic apomorphies shared with the North American form. Palaeochiropteryx,Archaeonycteris, and Hassianycteris are known principally from the famous Early/Middle Eocene ‘‘Grube Messel’’ deposits (MP 11) near Darmstadt, Germany. Hundreds of complete and partial skeletons of bats have been found at Messel, many preserved with stomach contents indicating that they were insectivores that had been foraging successfully just prior to death (Smith et al., 1979; Richter and Storch, 1980; Richter, 1987; Habersetzer et al., 1992, 1994). This unusual concentration of fossil bats that show no evidence for cause of death may have resulted from release of poisonous gas that overcame bats foraging over the surface of Lake Messel (Habersetzer and Storch, 1988; Habersetzer et al., 1992). Another possibility is that the bats died as a result of poisoning by toxic alkaloids from a bluegreen algal bloom (see Pybus et al. [1986] for a modern example). Palaeochiropteryx is by far the most common bat at Messel, accounting for almost 75% of all bat finds (Habersetzer and Storch, 1989; Habersetzer et al., 1992). More than 50 skeletons of Palaeochiropteryx are known from Messel, and two species are currently consists of continuously evolving lineages; therefore, separation between ancestor and descendant species (and the temporal range of each species) is completely arbitrary. In an effort to prevent this subjective element from confounding biostratigraphy, the boundaries of the MP units were not defined by the first and last appearance of taxa; accordingly, they are not true biostratigraphic zones (Schmidt-Kittler, 1987). Each MP level has a reference locality whose fauna functions as the type for the reference level (Schmidt-Kittler, 1987). The method used to order MP levels not correlated to marine stages is not clear. It appears to be based on faunal similarity and the evolutionary stage of the lineages within the respective faunas (Russel et al., 1982; Schmidt-Kittler, 1987). Ordering of the MP levels by radiometric ages, magnetostratigraphy, and the superposition of strata has not been possible (Schmidt-Kittler, 1987). Although the use of faunal evolutionary stage in developing the sequence of MP levels is subjective and of questionable validity, it is currently the only option for placing European fossil mammals in geologic time. recognized: Palaeochiropteryx tupaiodon Revilliod, 1917 (fig. 2) and Palaeochiropteryx spiegeli Revilliod, 1917 (Smith and Storch, 1981; Habersetzer and Storch, 1987). P.tupaiodon was the smaller of the two species, probably weighing 7–10 g and having an estimated wingspan of 24–28 cm (Habersetzer and Storch, 1987; Norberg, 1989). In contrast, P.spiegeli had an estimated wingspan of 26–30 cm and probably weighed 10– 18 g (Habersetzer and Storch, 1987; Norberg, 1989). In addition to the Messel material, isolated fragments referable to Palaeochiropteryx have also been reported from Sparnacian, Cuisian, Lutetian, and Bartonian (MP 10, 11–13, 16) deposits from elsewhere in Europe (Russell et al., 1973, 1982; Savage and Russell, 1983). Archaeonycteris is known from at least six skeletons from Messel, and two species are currently recognized based on this material: Archaeonycteris trigonodon Revilliod, 1917b (fig. 3) and A.pollex Storch and Habersetzer, 1988 (Russell and Sige´, 1970; Smith and Storch, 1981; Habersetzer and Storch, 1987; Storch and Habersetzer, 1988). A.trigonodon probably had a wingspan of 32–37 cm and may have weighed 17–27 g (Habersetzer and Storch, 1987; Norberg, 1989). A.pollex was slightly larger, with an estimated body weight of 30–35 g (Storch and Habsersetzer, 1988). In addition to the Messel material, fragmentary specimens referable to Archaeonycteris have been reported from Sparnacian, Cuisian, and possibly Bartonian deposits in France (MP 8–13, 16; Russell et al, 1973, 1982; Godinot, 1981; Savage and Russell, 1983; Schmidt-Kittler, 1987). This material includes at least one additional species, Archaeonycteris brailloni Russell, Louis, and Savage, 1973 (MP 8–9; Schmidt-Kittler, 1987). Another species, Archaeonycteris revilliodi Russell and Sige´, 1970, is based on a partial dentition from Messel. Always considered somewhat problematic, this form was transferred to Hassianycterididae as Hassianycteris?revilliodi by Habersetzer and Storch (1987). Smith and Russell (1992) subsequently reported discovery of a skull of revilliodi and confirmed that this species should be placed in Hassianycteris. Hassianycteris is known from at least 20 skeletons from Messel. Three species are 1998 7SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS currently recognized based on this material: Hassianycteris messelensis Smith and Storch, 1981; Hassianycteris magna Smith and Storch, 1981; and Hassianycteris revilliodi Russell and Sige´, 1970 (Smith and Storch, 1981; Habersetzer and Storch, 1988; Smith and Russell, 1992). Wingspans and weights have been estimated for only those species known from relatively complete skeletons (i.e., H.messelensis and H.magna).H. messelensis probably had a wingspan of 35– 40 cm and weighed 25–45 g (Habersetzer and Storch, 1987; Norberg, 1989). H.magna had a wingspan of 45–50 cm and may have weighed 65 g, making it the largest known Eocene bat (Habersetzer and Storch, 1987; Norberg, 1989). A fourth species, Hassianycteris joeli Smith and Russell, 1992, is known from a partial dentary with teeth from upper Ypresian (Early Eocene) deposits in Belgium (Smith and Russell, 1992). RELATIONSHIPS AND CLASSIFICATION OF EOCENE BATS: A HISTORICAL OVERVIEW Phylogenetic relationships among Icaronycteris,Palaeochiropteryx,Archaeonycteris, and Hassianycteris have been the subject of considerable debate, as have their relationships to extant lineages. It has long been thought that many Eocene fossils represent early members of extant microchiropteran families, but Icaronycteris,Palaeochiropteryx,Archaeonycteris, and Hassianycteris (often referred to as the ‘‘archaic’’ Eocene bats) have remained enigmatic. Most workers have regarded some or all of these forms as representatives of an early grade of chiropteran evolution, but opinions have differed concerning their relationships to each other, other Eocene taxa, Megachiroptera, Microchiroptera, and to extant microchiropteran superfamilies. This uncertainty has been reflected in classifications, which have varied considerably over the last few decades. Because ideas concerning relationships of Icaronycteris,Palaeochiropteryx,Archaeonycteris, and Hassianycteris developed in the context of a rich body of literature on early Tertiary bats, we review here (in chronological order) the history of classification and phylogenetic hypotheses regarding all Eocene chiropteran taxa, not just the four genera examined in the present study. The taxonomy and spelling of names are those of the original author(s) of each study. Question marks associated with taxonomic names to indicate uncertainty (e.g., Icaronycteris? menui) also reflect the usage of the original author(s). Twenty-four genera of bats are now recognized from the Eocene, although some were originally described based on material from younger deposits (table 1). Several early Tertiary bat fossils were reported in the literature prior to 1875, but most were isolated teeth or fragments of jaws referred to either Chiroptera, Rhinolophus,orVespertilio, the latter being a waste-basket taxon that once included most bats. Summaries of the history of fossil bat discoveries prior to 1875 were provided by Revilliod (1922) and Legendre and Sige´ (1983). Most modern bat classifications are based largely upon that of Dobson (1875: 345), who was the first to provide a comprehensive classification of extant bats ‘‘arranged according to their natural affinities.’’ Dobson (1875) briefly considered the origins of Chiroptera, although he did not explicitly discuss any of the fossil material known at the time. Accompanying Dobsons (1875) account was a figure ‘‘illustrating the affinities of the families and genera of Chiroptera, and probable lines of descent from ancestral forms . . .’’ (fig. 5). In this diagram, Dobson (1875) used Palaeochiroptera as a name for the group of largely unknown fossil bats that he presumed were ancestral to all extant bat lineages. Schlosser (1887) named Pseudorhinolophus and Vespertiliavus based on material from the Late Eocene to Early Oligocene Quercy phosphorite deposits in France (currently referred to MP 16–23; Crochet et al., 1981; Sige´ and Legendre, 1983; SchmidtKittler, 1987). Drawing on extensive comparisons with extant genera, Schlosser (1887) suggested a close relationship between Pseudorhinolophus and Rhinolophus and noted several points of similarity between Vespertiliavus, species of Vespertilio (many now placed in other genera and even other families; see below), and Taphozous. He subsequently concluded that Pseudorhinolophus should be grouped with rhinolophids and 8 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY Fig. 1. Chart showing geologic time scale, standard geologic ages based on marine strata, and the correlation of land mammal ages and levels. Data complied from Fahlbusch (1976), Russell et al. (1982), Savage and Russell (1983), Berggren et al. (1995), Woodburne and Swisher (1995), and McKenna and Bell (1997). Correlations of MP levels to marine strata of known age are based on the following: MP 7 (Hooker, 1991; Woodburne and Swisher, 1995); MP 10, 13, 14, and 15 (Russell et al., 1982; Aubry, 1983); and MP 21 (Hooker, 1992). Abbreviations: EU., European; Ma., milleannus, millions of years before present; N.A., North American. For a definition of MP levels, see footnote 1 on page 5. a Ages according to Fahlbusch (1976). b Ages according to Savage and Russell (1983). c These MP reference levels are not correlated to marine strata; instead, they are placed in an approximate chronological order (based on Russell et al. [1982] and Schmidt-Kittler [1987]) between MP levels that can be correlated to marine strata. Vespertiliavus with vespertilionids. Schlosser (1887) also described fragmentary material that he referred to Rhinolophus sp. and to a later Tertiary species of Rhinolophus described by Filhol (1872). Schlosser (1887) also discussed an isolated humerus that he suggested was similar to that of rhinolophids. Weithofer (1887) discussed new material of Pseudorhinolophus and suggested modifications of Schlosser’s (1887) description and allocation of this taxon. Based on comparisons with extant material, Weithofer (1887) concluded that Pseudorhinolophus was more closely related to Phyllorhina ( 5 Hipposideros) than to Rhinolophus.Hereferred the isolated humerus described by Schlosser (1887) to ‘‘Taphozous(?),’’ noting that it was clearly different from the typical rhinolophid form. Weithofer (1887) also described several new taxa based on craniodental material from the Quercy deposits, including Alastor (which he associated with Pseudorhinolophus and Phyllorhina), Necromantis (which he concluded was related to 1998 9SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS Fig. 2. Palaeochiropteryx tupaiodon (SMF ME 10) from Messel, Germany. From Habersetzer and Storch (1987: fig. 2). Photo by E. Pantak (Senkenbergmuseum). Phyllostomidae), and a fossil species referred to ‘‘Rhinolophus(?).’’ Recent collections have suggested that Necromantis is restricted to the MP 17 fauna (Sige´ and Legendre, 1983; Schmidt-Kittler, 1987). Winge (1892 [translated into English in Winge, 1941]) discussed the status of Pseudorhinolophus and Alastor, noting that although some authors had treated these as genera separate from Phyllorhina ( 5 Hipposideros), few had noted any significant differences. He concluded that Palaeophyllophora clearly belonged to the same group as Pseudorhinolophus, although Palaeophyllophora is somewhat more specialized. Winge (1892) also discussed an undescribed, almost complete skull of Vespertiliavus from the collections at Copenhagen. Several features of this specimen led him to conclude that Vespertiliavus is a relatively specialized member of Emballonuridae, closely related to Taphozous. He noted that this conclusion was not surprising given the taxonomic history of species previously compared with Vespertiliavus. Citing Schlossers (1887) original description of Vespertiliavus, Winge (1892) observed that the ‘‘vespertilionid’’ that Schlosser thought was most similar to 10 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY Fig. 3. Archaeonycteris trigonodon (SMF 80/1379) from Messel, Germany. From Habersetzer and Storch (1987: fig. 3). Photo by E. Pantak (Senkenbergmuseum). Vespertiliavus—Vespertilio alecto—was considered by Dobson (1878) to be conspecific with Emballonura monticola. Winge further noted that the isolated humerus referred to ‘‘Taphozous(?)’’ by Weithofer (1887) undoubtedly also represents Vespertiliavus. Meschinelli (1903) described Archaeopteropus based on a skeleton with a poorly preserved skull from late Oligocene deposits 1998 17SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS Palaeochiropterygidae (including Palaeochiropteryx and ?Cecilionycteris). Both of these families were left in ‘‘Superfamily uncertain’’ within Microchiroptera (Romer, 1966: 382). Paradoxonycteris was referred to Chiroptera incertae sedis (Romer, 1966). Jepsen (1966) described Icaronycteris from a single, beautifully preserved skeleton from Fossil Basin in the Green River Formation of Wyoming. He placed this form in the new family Icaronycteridae within Microchiroptera. Jepsen discussed other putative bats from early Tertiary deposits in North America, but did not make any detailed comparisons with Eocene bats from Europe. Sige´ (1968) described the bats of an Early Miocene fauna from Bouzigue, France, including a new species of Pseudorhinolophus. He reviewed the contents and stratigraphic range of Pseudorhinolophus and presented a revised diagnosis of this taxon, which he considered to represent a subgenus of Hipposideros. He suggested several possibilities for relationships among Pseudorhinolophus, Brachyhipposideros (a subgenus known from Oligocene and Miocene deposits), and modern species of Hipposideros, but he reached no definitive conclusions other than that these forms are very closely related. Jepsen (1970) published detailed stereophotographs of Icaronycteris in an essay concerning the evolution of bats and powered flight. However, the text of his publication contained little information on Icaronycteris that was not available in the original description (Jepsen, 1966), and no comparisons were drawn between Icaronycteris and Eocene bats from Europe. Concerning the relationships of Icaronycteris to other bats, Jepsen (1970: 40) noted only that ‘‘Icaronycteris index as a species may have been directly ancestral to all or to some living microbats or megabats or to none of our contemporary chiropts.’’ In a short paper on the classification of bats, Koopman and Jones (1970) recognized three families of Eocene bats: Icaronycteridae (including only Icaronycteris), Archaeonycteridae (Archaeonycteris), and Palaeochiropterygidae (Palaeochiropteryx and Cecilionycteris, the latter of which was placed incertae sedis). These families were listed under the heading ‘‘Fossil Chiroptera of uncertain status’’ (Koopman and Jones, 1970: 28). Paleunycteris and Paradoxonycteris were referred to Chiroptera incertae sedis, and several Eocene taxa were placed in modern families: Vespertiliavus in Emballonuridae, Necromantis in Megadermatidae, and Stehlinia in Vespertilionidae (Koopman and Jones, 1970). Palaeophyllophora and Paraphyllophora were placed in their own tribe Palaeophyllophorini in Hipposiderinae (Koopman and Jones, 1970). Archaeopteropus was placed in Archaeopteropodinae within Megachiroptera. Slaughter (1970) discussed evolution of the chiropteran dentition, drawing conclusions about the affinities of many fossil taxa based on their dental morphology. Slaughter (1970: 59) observed that the dentition of Icaronycteris is ‘‘more like the ancestral condition than any other known chiropteran.’’ He went on to note that Cecilionycteris,Palaeochiropteryx, and Archaeonycteris have slightly more derived dentitions, but did not associate them with any extant lineages. He observed that Necromantis is clearly a megadermatid, although somewhat less derived than extant members of the family. Slaughter placed Pseudorhinolophus on the lineage leading to modern Hipposideros, and recognized Palaeophyllophorinae as close relatives of hipposiderines. The ‘‘Eocene Rhinolophus’’ were placed at the base of the lineage leading to modern Rhinolophus, palaeophyllophorines, and hipposiderines (Slaughter, 1970: fig. 5), with Nycteris shown as the sister-group to this rhinolophid clade. He also recognized Vespertiliavus as the oldest member of Emballonuridae, noting that the prototypic dentition for the superfamily Emballonuroidea (which he defined as including noctilionids) would be like that of Vespertiliavus except for the form of the hypocone. Finally, Slaughter noted that Stehlinia appears to be intermediate between ancestral vespertilionids and the lineage leading to all vespertilionids except miniopterines and murinines. Russell and Sige´ (1970) were among the first to present detailed comparisons of the Messel bats with extant taxa. Their study concluded with an updated classification of the archaic Eocene bats and emended diag- 18 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY Fig. 6. Russell, Louis, and Savage’s hypothesis of phylogeny and divergence times of Eocene bats (redrawn from Russell et al., 1973: fig. 11). The original caption for this figure read ‘‘Hypothetical relationships between the earliest European bats.’’ noses for each taxon. They placed the archaic Eocene genera in a new superfamily Palaeochiropterygoidea within Microchiroptera. Two families were recognized within this group: Palaeochiropterygidae and Icaronycterididae. Icaronycterididae was defined as including Icaronycteris and possibly Archaeopteropus. In contrast, Palaeochiropterygidae was divided into two subfamilies, Palaeochiropteryginae (including Palaeochiropteryx and Cecilionycteris) and Archaeonycteridinae (containing only Archaeonycteris). Sige´ (1971) described an isolated forelimb from Stampien deposits in France (MP 21– 24; Russell et al., 1982; Schmidt-Kittler, 1987) and referred it to Tadarida sp. Although this specimen may represent Cuvierimops, lack of associated dental material has prevented a confirmed identification. Hand (1990) referred this specimen to Tadaridinae indet. (Tadaridinae was named by Legendre [1984] in a review of extant molossids). Russell et al. (1973) described fragmentary dental remains from the Early Eocene of France (MP 8–9; Schmidt-Kittler, 1987) that they referred to new species of Icaronycteris? and Archaeonycteris as well as a new genus, Ageina. Based on dental morphology, they hypothesized that Icaronycteris?menui was closely related to Palaeochiropteryx.Archaeonycteris was recognized as a distinct yet related lineage, while Ageina was placed on a distant branch of uncertain affinities (fig. 6; Russell et al., 1973: fig. 11). Although classification was not explicitly discussed, they placed Icaronycteris?menui in Palaeochiropterygoidea: Icaronycteridae?, 1998 19SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS Fig. 7. Smith’s hypotheses of bat relationships (redrawn from Smith, 1976: fig. 1). The original caption for this figure read ‘‘A, cladogram of the generally accepted view of chiropteran phylogeny with the Microchiroptera dervied from a common emballonuridlike ancestry. B, an alternative proposal for chiropteran evolution with several microchiropteran lineages being derivied, independently, from a world-wide paleochiropteran grade and the Megachiroptera (Pteropodidae) derived either separately from an insectivorous stock or from the paleochiropteran grade. a, Emballonuroidea; b, Rhinolophoidea; c, Phyllostomatoidea; d, Vespertilionoidea.’’ Note that Smith (1976) used ‘‘Myzapodidae’’ instead of ‘‘Myzopodidae,’’ and ‘‘Phyllostomatoidea’’ instead of ‘‘Phyllostomoidea’’ ( 5 Noctilionoidea). Archaeonycteris brailloni in Archaeonycterinae (family not specified), and Ageina in ‘‘Family uncertain’’ (Russell et al., 1973: 35). Sige´ (1974) reviewed the available material of Stehlinia and compared it to other fossil and extant taxa. Based on his analysis, Sige´ recognized Nycterobius,Paleunycteris, and Revilliodia as junior synonyms of Steh- 20 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY linia, and concluded that Stehlinia was best placed in the family Kerivoulidae within Vespertilionoidea. He compared Stehlinia with ‘‘les pale´ochiropte´rygoı¨de´s’’ but did not comment on contents or classification of this group. Smith (1976) discussed Eocene fossil bats in a review of phylogenetic relationships and patterns of evolutionary radiation of extant bats (fig. 7). Apparently drawing on Dobson’s (1875) concept, Smith (1976: 52–53) used the informal name ‘‘Paleochiroptera’’ to collectively refer to Icaronycteris,Palaeochiropteryx,Archaeonycteris,Cecilionycteris, and Ageina. He perceived this group as representing ‘‘a world-wide paleochiropteran grade in early to middle Eocene times . . . [that] was primitive and generalized in most respects.’’ He suggested that extant Microchiroptera could be easily derived from the paleochiropteran grade, and that this divergence probably occurred in the Paleocene. He clearly did not think that any of the Eocene taxa listed above were ancestral to extant microchiropterans. Smith further suggested that Archaeopteropus was probably also derived from the paleochiropteran grade, although he questioned the megachiropteran affinities of this form. Smith (1976: 54) observed that ‘‘The distinctness and marked departure of megachiropteran dentition from that of the Microchiroptera, as well as from known Tertiary fossils, suggests to me that this group of bats had their origin much earlier in the paleochiropteran grade or perhaps . . . separately from an insectivorous ancestral stock.’’ Smith (1976) also discussed other Eocene fossils that he thought could be referred to extant families. These included Vespertiliavus (referred to Emballonuridae); Palaeophyllophora,Paraphyllophora,Rhinolophus, Pseudorhinolophus, and Hipposideros (Rhinolophidae); Necromantis and Provampyrus (Megadermatidae); and Stehlinia and Nycterobius (Vespertilionidae). The presence of modern families (and even genera) in Middle and Late Eocene deposits was interpreted as additional evidence for an early divergence of microchiropterans from the paleochiropteran stock (Smith, 1976). Sige´ (1977) discussed the fossil record of Eocene bats in a review of Paleogene mammalian faunas of Europe. He referred to the Messel bats as ‘‘chiropte`res pale´ochiropte´rygoı¨de´s,’’ including within this group Icaronycterididae and Palaeochiropterygidae. The latter family contained two subfamilies: Archaeonycteridinae and Palaeochiropteryginae. Vespertiliavus was referred to Emballonuridae, Stehlinia to Vespertilionoidea (Kerivoulidae), Hipposideros (Pseudorhinolophus) and Palaeophyllophora to Rhinolophoidea (Hipposideridae), and Necromantis to Megadermatidae. In the absence of contradictory evidence, Sige´ suggested that palaeochiropterygines, emballonuroids, vespertilionoids, and rhinolophoids might have originated from unspecified endemic European groups. Based on its supposed ‘‘denture apparemment insectivore’’ (insectivorous dentition), Sige´ (1977: 185) interpreted Archaeopteropus as a form intermediate between megachiropterans and icaronycterids. In a review paper concerning the evolution of bat flight, Smith (1977) noted that one of the ‘‘five monophyletic superfamilies’’ of Microchiroptera was Palaeochiropterygoidea, which included Palaeochiropterygidae and Icaronycteridae. While Russell and Sige´ (1970) had tentatively referred Archaeopteropus to Icaronycteridae, Smith (1977) noted that he had examined the only known specimen of Archaeopteropus and had found no justification for this arrangement. Accordingly, he restricted Icaronycteridae to only Icaronycteris. To our knowledge, Smith was the first to suggest that Icaronycteris,Palaeochiropteryx,Archaeonycteris,Cecilionycteris, and Ageina formed a clade (rather than a grade) distinct from other microchiropteran bats. Barghoorn (1977) described two new skulls of Vespertiliavus from the Quercy Phosphorites in southern France. Based on a cladistic analysis of cranial characters in extant and fossil emballonurids and several outgroups, Barghoorn concluded that Vespertiliavus is the sister-group of a clade comprised of Taphozous and Saccolaimus. Butler (1978) reviewed the evolutionary history of bats in Africa and provided a classification of fossil forms known from that continent. In this context, he placed Vampyravus in Microchiroptera incertae sedis, ex- 1998 21SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS Fig. 8. Van Valen’s (1979) hypothesis of bat relationships (redrawn from Van Valen, 1979: fig. 1). The original caption for this figure read ‘‘Phylogeny of the known families of bats . . . Infraorders, superfamilies, and families of the Microchiroptera are separated by solid, dashed, and dotted lines respectively.’’ The numbers refer to taxa discussed in an appendix; we reproduce these here because they are informative with respect to Van Valen’s unconventional concepts of group membership: 1, Megachiroptera, Pteropodidae; 2, Microchiroptera, undiscovered family; 3, Vespertilionia, Vespertilionoidea, Natalidae; 4, Recent Natalidae; 5, Vespertilionidae; 6, Recent Vespertilionidae; 7, Mystacinidae; 8, Recent Mystacinidae; 9, Molossidae; 10, Phyllostomatia, Rhinopomatoidea, undiscovered Family; 11, Emballonurid–rhinolophoid stem; 12, Rhinolophoidea; 13, Megadermatidae; 14, Nycteridae; 15, Rhinolophidae; 16, Rhinopomatidae; 17, Emballonurid–noctilionoid stem; 18, Emballonuridae (family should perhaps extand as far back as 11); 19, Recent Emballonuridae; 20, Craseonycterididae; 21, Noctilionoidea, perhaps early Mormoopidae; 22, Noctilionidae; 23, Mormoopidae; 24, Recent Mormoopidae; 25, Phyllostomatidae; 26, Recent Phyllostomatidae; 27, Desmodontidae. Spelling of group names follows those of the original author. plicitly rejecting the conclusions of Schlosser (1911) and Smith (1976). Van Valen (1979) published the first cladistic assessment of higher-level bat relationships (fig. 8). This study was based on diverse morphological characters, although specific methods of analysis were not described and a taxon–character matrix was not included. The focus of Van Valen’s study was extant bats, but he also discussed some features of the better known fossil forms as reported in the literature. He proposed the new suborder Eochiroptera for the archaic Eocene bats, within which he recognized one family (Palaeochiropterygidae). The latter group was described as ‘‘including Archaeonycter(id)idae and probably Icaronycter(id)idae and Archaeopteropodinae as subfamilies’’ (Van Valen, 1979: 109). Van Valen indicated that Eochiroptera was ancestral to all extant bats including Megachiroptera and Microchiroptera. He noted that extant megachiropterans were probably derived from persistent eochiropterans such as Archaeopteropus. The 22 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY criteria used to lump forms ancestral to different clades into one group (Eochiroptera) appears to have been perceived adaptive similarity and their shared retention of numerous primitive traits. Van Valen (1979: 109) acknowledged that many of the taxa in his classification were paraphyletic, which he noted ‘‘reflects the progressive evolution of grades of adaptation.’’ Sige´ and Russell (1980) provided detailed descriptions and comparisons of Palaeochiropteryx tupaiodon and Cecilionycteris, and described a new genus Matthesia based on dental material from Geiseltal, ‘‘Grube Cecilie’’ (MP 13; Franzen and Haubold, 1987). All of these forms were placed in Eochiroptera: Palaeochiropterygidae: Palaeochiropteryginae by Sige´ and Russell (1980). Smith and Storch (1981) described the new genus Hassianycteris from Messel (MP 12; Schmidt-Kittler, 1987). Hassianycteris was described as differing significantly from palaeochiropterygoids in possessing an advanced degree of dental reduction and other derived dental and osteological features. Smith and Storch (1981: 164) concluded that Hassianycteris ‘‘is more closely associated with the emballonuroid/rhinolophoid section of the Microchiroptera than to any other group.’’ As a result, they referred Hassianycteris to Microchiroptera incertae sedis rather than placing it in Palaeochiropterygoidea. Following many earlier authors, Smith and Storch (1981) recommended removal of Archaeopteropus from Palaeochiropterygoidea and subsequent placement in Pteropodidae based on morphological evidence that Archaeopteropus is an early megachiropteran (for details of this argument, see Habersetzer and Storch, 1987). In doing so, Smith and Storch explicitly restricted Palaeochiropterygoidea to Palaeochiropteryx,Archaeonycteris,Icaronycteris,Cecilionycteris,Ageina, and Matthesia. They (1981: 163) noted the following concerning possible relationships of Palaeochiropterygoidea as thus defined: While we would not now go so far as including these bats in the family Vespertilionidae Gray 1821, they may well share a close sister-group relationship with the superfamily Vespertilionoidea or they may be placed as the sister-group of a larger one including the Phyllostomatoidea and Vespertilionoidea. Legendre and Sige´ (1983) went back to one of the first fossil bats discussed in the scientific literature—Cuvier’s (1822) ‘‘Vespertilion de Montmartre’’—and found that additional preparation of the original specimen (a partial skeleton) revealed features diagnostic of Molossidae. This form, which was variously referred to Vespertilio parisiensis,Serotinoides antiquus,‘‘Vespertilio’’ serotinoides, and cf. Tadarida sp. by previous 19th-century authors, was named Cuvierimops by Legendre and Sige´ (1983). It is currently thought to be Late Eocene (Priabonian) in age (MP 19; Russell et al., 1982; Schmidt-Kittler, 1987). Legendre and Sige´ (1983) suggested that Cuvierimops was possibly ancestral to a large complex including all extant molossids with the exception of Tadarida and Mormopterus. Sige´ and Legendre (1983) reviewed the fossil record of bats in different depositional environments and regions in the Mediterranean basin. They referred Icaronycteris,Palaeochiropteryx,Cecilionycteris,Matthesia, Archaeonycteris, and Ageina to Eochiroptera. Hassianycteris was described as an early microchiropteran not clearly related to any of the extant superfamilies. Several additional Eocene forms were referred to extant families, including Vespertiliavus,Hipposideros (Pseudorhinolophus), Palaeophyllophora, Necromantis,Rhinolophus,Stehlinia,Cuvierimops, and cf. Tadarida. Although family affinities were not specified, these references were clearly intended to follow past classifications of these taxa. Ostrander (1983) described Chadronycteris based on a maxillary fragment with teeth from what are now considered to be Late Eocene (Chadronian) deposits in Nebraska. Based on comparisons with Stehlinia (referred to Kerivoulidae by Sige´, 1974) and ‘‘Palaeochiropterygidae,’’ Ostrander referred Chadronycteris to Vespertilionoidea: Kerivoulidae. Gupta (1984) discussed bat phylogeny and relationships of Icaronycteris. Apparently failing to draw any distinction between symplesiomorphy and synapomorphy, Gupta suggested that Icaronycteris may be a megachiropteran based on several shared primitive features (e.g., presence of a claw on the index finger). Gupta (1984: 42) also pro- 1998 23SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS Fig. 9. Cladogram of interfamilial relationships inferred from auditory characters (redrawn from Novacek, 1980a: fig. 1). The original caption for this figure read ‘‘Wagner tree resulting from analysis of 18 auditory characters for Recent chiropteran families. Craseonycteridae and Myzopodidae are omitted. Horizontal bars bridging two or more branches are synapomorphies; short black bars are independently derived characters under this arrangement; striped bars represent character state reversals.’’ posed an unusual hypothesis concerning the origin of bats: The close similarity in the structure of the patagium of Pterosaurs and of bats, and the presence of hair in both, can compel us to think that the bats might have evolved from the Pterosaurs with modifications of mammalian characters during 12 million years [of the] Paleocene, the period during which no fossils [of pterosaurs or bats] have so far been reported. None of these ideas have been accepted by subsequent authors because all published data sets indicate that bats are therian mammals (e.g., Novacek, 1980a, 1986, 1990; Novacek and Wyss, 1986; Simmons, 1993a, 1994, 1995). Pterosaurs are almost universally regarded as archosaurian diapsid reptiles that are more closely related to crocodiles and birds than to mammals (e.g., Padian, 1984, 1987; Gauthier, 1986). Hill and Smith (1984) provided a classification of all chiropteran genera in their book on the natural history of bats. They recognized three families in Palaeochiropterygoidea: Palaeochiropterygidae (including Palaeochiropteryx,Cecilionycteris, and Matthesia), Archaeonycteridae (including Archaeonycteris and Ageina), and Icaronycteridae (including only Icaronycteris). Paradoxonycteris and Hassianycteris were left as ‘‘Family incertae sedis’’ within Microchiroptera (Hill and Smith, 1984: 221). Following Smith and Storch (1981), Hill and Smith (1984) suggested that Palaeochiropterygoidea appeared to have affinities with Vespertilionoidea, while Hassianycteris might be associated with Emballonuroidea and Rhinolophoidea. Many Eocene and Oligocene fossils were referred to extant families by Hill and Smith (1984). They placed Archaeopteropus in its own subfamily Ar- 24 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY chaeopteropodinae within Pteropodidae, Vespertiliavus in Emballonurinae within Emballonuridae, Necromantis and Provampyrus in Megadermatidae, Pseudorhinolophus in Hipposiderinae within Hipposideridae, Palaeophyllophora and Paraphyllophora in Palaeophyllophorinae within Hipposideridae, and Stehlinia in Vespertilioninae within Vespertilionidae. Legendre (1985) reviewed the fossil record of molossids and reevaluated the affinities of Wallia, a taxon originally described by Storer (1984) as a proscalopid insectivore from the Uintan (Middle Eocene) Swift Creek Local Fauna of Saskatchewan, Canada. Legendre concluded that Wallia probably represents the oldest known molossid, and he placed both Wallia and Cuvierimops in Tadaridinae, the subfamily to which all Tertiary fossil molossids have been referred (Hand, 1990). Sige´ (1985) focused on fossil bats from Late Eocene deposits of the Fayum area, Egypt, and redescribed and discussed the status of Vampyravus Schlosser, 1910 and Provampyrus Schlosser, 1911. He concluded that these two taxa are objective synonyms based on a single specimen, a well-preserved humerus. Sige´ (1985) concluded that Vampyravus was too poorly known to be assigned to any family, but suggested that morphology of the holotype indicated affinities with either hipposiderids, natalids, or phyllostomids. Sige´ also described a new taxon based on craniodental material, Philisis, which he placed in its own family, Philisidae. Based on comparisons with fossil and extant forms, Sige´ placed Philisidae within Vespertilionoidea and suggested that it was more closely related to Vespertilionidae sensu lato than to Molossidae (fig. 12). Speculating that Vampyravus and Philisis might eventually be shown to be conspecific if more complete material were to be discovered, Sige´ (1985) hypothesized that Vampyravidae (which would replace Philisidae as the correct family name in this case) would prove to be more closely related to Natalidae than to Vespertilionidae or Molossidae. In each of Sige´’s phylogenetic trees (fig. 12), Eochiroptera was shown as occupying the basal branch. It is not clear, however, whether this was meant to imply monophyly of Eochiroptera; comments in the text suggest that Sige´ included Eochiropterato represent the basal stock of Chiroptera. Hora´cek (1986) discussed relationships of Stehlinia and Kerivoula and concluded (contra Sige´, 1974) that these taxa were not closely related. Instead, Hora´cek suggested that Stehlinia might be related to Miniopterinae. Mein and Tupinier (1986) briefly reviewed the early Tertiary record of bats in a short paper on the evolution of echolocation systems. They followed Van Valen (1979) in considering Icaronycteris,Archaeonycteris, and Palaeochiropteryx as representatives of Eochiroptera, while referring other fossil taxa to extant superfamilies or families. Mein and Tupinier listed Vespertiliavus in Emballonuridae, Stehlinia in Vespertilionoidea, and referred Necromantis and possibly Provampyrus to Megadermatidae. Gingerich (1987) described Wyonycteris based on a partial dentary and isolated teeth from Late Paleocene (Clarkforkian) deposits of the Willwood Formation of Wyoming. He referred Wyonycteris to Palaeochiropterygidae on the basis of dental similarities, but subsequent authors (e.g., Habersetzer et al., 1994; Hand et al., 1994) have questioned the chiropteran affinities of this form (see footnotes to table 1). Gingerich also described teeth referred to cf. Icaronycteris from the Willwood Formation. Habersetzer and Storch (1987) discussed in detail the classification and functional morphology of Paleogene bats, concentrating on the fossil bats from Messel. They argued against usage of Palaeochiropterygoidea and Eochiroptera, recognizing that each was probably paraphyletic even if Archaeopteropus was removed. Instead, they recognized and provided revised diagnoses for three families placed in Microchiroptera incertae sedis: Archaeonycterididae (including Archaeonycteris and Icaronycteris), Palaeochiropterygidae (Palaeochiropteryx), and a new family, Hassianycterididae (Hassianycteris, including ‘‘Archaeonycteris’’ revilliodi Russell and Sige´, 1970). The relationships of Cecilionycteris,Ageina, and Matthesia were not addressed. Novacek (1987) focused on phylogenetic relationships of the best known Eocene taxa (Icaronycteris and Palaeochiropteryx) rather 1998 25SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS Fig. 10. Relationships of selected taxa inferred from features of fetal membrane structure and development (redrawn from Luckett, 1980a: fig. 5). The original caption for this figure read ‘‘Character state distribution of fetal membrane and reproductive features of Chiroptera . . .’’ than on issues of classification. Following a detailed review of a wide array of morphological features, Novacek (1987) concluded that both Icaronycteris and Palaeochiropteryx are more closely related to extant microchiropterans than to extant megachiropterans (fig. 14). He concluded that Icaronycteris and probably Palaeochiropteryx are outgroups of all Recent families of Microchiroptera, but could not rule out the possibility that one or both of these Eocene taxa might have special affinities within Microchiroptera (i.e., might be more closely related to one or more extant families than to others). Novacek (1987: 15–16) further concluded that . . . the Palaeochiropterygoidea seems merely an artificial convention to group several early microchiropterans whose relationships with modern families of this suborder remain poorly known . . . Icaronycteris and Palaeochiropteryx are more accurately designated as Microchiroptera incertae sedis. Moreover, there seems no justification for a formal designation of an ‘‘ancestral’’ group (e.g., Eochiroptera sensu Van Valen, 1979) to distinguish Icaronycteris,Palaeochiropteryx, and other Eocene bats from the Recent chiropteran suborders. In a classification of fossil vertebrates, Carroll (1988) named a new superfamily for the archaic Eocene bats, Icaronycteroidea, which he placed within Microchiroptera. He included two families in Icaronycteroidea: Icaronycteridae (Icaronycteris) and Palaeochiropterygidae (Archaeonycteris,Palaeochiropteryx,Cecilionycteris, and Matthesia). In concept and contents, Icaronycteroidea is identical to Palaeochiropterygoidea as recognized by most previous authors. Carroll gave no justification for this apparently unnecessary name change, and no subsequent authors have used Icaronycteroidea. Carroll additionally provided a classification for other Eocene bats, placing Vespertiliavus in Emballonuridae, Necromantis in Megadermatidae, Rhinolophus in Rhinolophidae, Hipposideros,Palaeophyllophora, and Paraphyllophora in Hipposideridae, ?Vampyravus in Phyllostomidae, and Stehlinia in Vespertilionidae. He also placed Archaeopteropus in Pteropodidae. Robbins and Sarich (1988) used protein electrophoresis and immunological distance data to address relationships among extant Emballonuridae. They also considered morphological data presented by Barghoorn (1977), and included a brief discussion of the affinities of Vespertiliavus. Robbins and Sarich concluded by assigning Vespertiliavus to a new tribe, Vespertiliavini, within the subfamily Taphozoinae. Sige´ (1988) described the fossil bats from the Marinesian ( 5 Robiacian; MP 16) Le Bretou fauna from the Quercy Phosphorites in France. He referred Vespertiliavus to Emballonuridae, and Palaeophyllophora and Hipposideros (Pseudohipposideros) to Hipposideridae. Sige´ (1988: 93) also reported the presence of a new form that he referred to ‘‘Rhinolophoidea, sp. indet.’’ Storch and Habersetzer (1988) described a new species of Archaeonycteris (A.pollex) based on two skeletons from Messel, and compared this taxon with other Eocene bats. 26 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY They also reported the discovery of a new skull that confirmed Habersetzer and Storch’s (1987) suggestion that ‘‘Archaeonycteris’’ revilliodi should be referred to Hassianycteris rather than to Archaeonycteris or Archaeonycterididae. Habersetzer and Storch (1989) discussed ecology and echolocation abilities of Eocene bats from Messel as inferred from wing and cochlear morphology. Based on comparisons with various extinct and extant bats, Habersetzer and Storch (1989: 214) noted that Archaeonycteris trigonodon Revilliod 1917 (family Archaeonycterididae) is an archaic species.... Palaeochiropteryx tupaiodon Revilliod 1917, and P. spiegeli Revilliod 1917, (family Palaeochiropterygidae) are small specialized species....Hassianycteris messelensis Smith and Storch 1981, H.magna Smith and Storch 1981, and H.revilliodi (Russell and Sige 1970) (Family Hassianycterididae) are the most advanced species in dental and skeletal features. Habersetzer and Storch (1989) argued against Van Valen’s (1979) concept of Eochiroptera, which was originally defined as including all of the taxa mentioned in the excerpt above. They noted that Eochiroptera as conceived by Van Valen would be a group of primitive and unspecialized species that ultimately gave rise to both Megachiroptera and Microchiroptera. Noting that the Messel bat fauna includes ecologically diverse forms that are all ‘‘completely developed’’ microchiropterans, Habersetzer and Storch (1989: 231) suggested that Eochiroptera was a misleading concept, at least when applied to the Messel bats. Sige´ (1990) and Sudre et al. (1990) described new species of Vespertiliavus from Stampian (MP 25) and early Bartonian (MP 14) deposits of France, respectively. Both retained Vespertiliavus in Emballonuridae. Sige´ additionally described a new species of Hipposideros (subgenus Pseudorhinolophus), a new species of Stehlinia (which he referred to Natalidae sensu Van Valen, 1979), and a new genus Vaylatsia. All four of the new taxa described by Sige´ (1990) were based on fragmentary dentitions and isolated postcranial elements from the Quercy Phosphorites. Vaylatsia was referred to Hipposideridae based on dental and humeral characters, but Sige´ (1990: 1132) noted that it ‘‘probably represents the stem group of the genus Rhinolophus.’’ Sige´ (1991a) later described the genus Dizzya from the Early Eocene of Chambi in northern Africa. Following broad comparisons with extant and extinct forms, he referred Dizzya to Vespertilionoidea: Philisidae. A specimen referred to Rhinolophoidea gen. and sp. indet. was also described, and Sige´ went on to extensively discuss the biostratigraphic, biogeographic, and paleoecological implications of the bat fauna from the African Eocene. In a final note, Sige´ discussed usage of the name Eochiroptera. Contra Habersetzer and Storch (1987, 1989) and Novacek (1987), he (1991a: 372) argued that this taxon ‘‘est pre´fe´re´ par logique, commodite´, et efficacite´a` celui super-familial des Palaeochiropterygoidea inclus dans les Microchiroptera.’’ Sige´ (1991a: 373) argued that recognizing the ‘‘taxon-grade’’ Eochiroptera is a simple, operational approach to classifying organisms of uncertain affinities that share a large number of primitive characters. Expressing the unusual opinion that Microchiroptera must itself be considered paraphyletic (‘‘l’unite´ sub-ordinale Microchiroptera, qui . . . doit eˆtre clairment perc¸ue comme un rassemblement paraphyle´tique, mais se valeur empirique. . ..’’; Sige´, 1991a: 373), he emphasized the empirical value of formally grouping taxa based on their perceived grade of evolution. He argued that such a classification system should perhaps be maintained for the chiropteran suborders (to indicate their adaptive level in the continuum of bat evolution), even if it becomes possible to diagnose strongly supported, monophyletic superfamilies including some of the archaic Eocene taxa. Sige´ (1991b) discussed morphology of the deciduous dentition of Eocene bats and presented a figure summarizing relationships and stratigraphic ranges (fig. 15). He indicated that five extant lineages (Megachiroptera, Phyllostomatoidea [ 5 Noctilionoidea], Rhinolophoidea, Emballonurioidea, and Vespertilionoidea) and three extinct lineages (Hassianycterididae, Archaeonycterididae, and Palaeochiropterygidae) were in existence in the Eocene. He placed Archaeonycterididae and Hassianycterididae in Eochiroptera, 1998 33SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS Fig. 16. Two alternative hypotheses of interfamilial relationships suggested by Novacek (redrawn from Novacek, 1991: figs. 6, 7). The original caption for Aread ‘‘ This cladogram . . . is based on discussions in Koopman (1984). The position of Palaeochiropteryx . . . follows Novacek (1987). Note the remote branch position of Rhinopomatidae.’’ the caption for Bread ‘‘as in [the previous figure] modified by relocation of Rhinopomatidae as a sister taxon of Rhinolophoidea (following Pierson, 1986).’’ together with Myzopodidae, Thyropteridae, and Furipteridae, which Van Valen (1979) reduced in rank to subfamily level. Novacek (1980a) and Luckett (1980a) published cladograms of bat relationships based on analyses of single organ systems. Novacek focused on morphology of the auditory region; Luckett concentrated on morphology and development of fetal membranes. Primarily because the tree derived from auditory characters (fig. 9) differed significantly from all previous hypotheses of 34 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY Fig. 17. Two alternative hypotheses of yinochiropteran relationships based on hyoid musculoskeletal morphology (redrawn from Griffiths et al., 1992: figs. 10, 11). Both trees were equally parsimonious in the context of the hyoid data presented by Griffiths and Smith (1991) and Griffiths et al. (1992), but these authors preferred tree A on the basis of hypothesized patterns of transformation in two characters. higher-level relationships, Novacek (1980a) warned against using that cladogram (or any other derived from a single organ system) as a basis for a new phylogenetic reconstruction or classification. Luckett’s (1980a) study was unable to resolve many relationships (fig. 10), but provided some support for monophyly of Noctilionoidea and a close relationship between Megadermatidae and a clade containing Vespertilionidae and Thyropteridae. Monophyly of Vespertilionidae sensu Koopman and Jones (1970) was apparently assumed by both Novacek (1980a) and Luckett (1980a). Eisenberg (1981) published a phylogeny of bats in his book on mammalian radiations. Eisenberg’s (1981: 147) tree (fig. 11) differed only slightly from that presented by Smith (1976; see fig. 7). Like Smith (1976), Eisenberg depicted each of the microchiropteran superfamilies as monophyletic. However, Eisenberg placed Emballonuroidea, Rhinolophoidea, and Phyllostomatoidea ( 5 Noctilionoidea) together in an unresolved clade with Vespertilionoidea as the sister-group, rather than identifying Emballonuroidea and Rhinolophoidea as sister-taxa as had Smith (1976). Gopalakrishna and Chari (1983) described fetal membrane development in Miniopterus, a taxon generally placed in its own subfamily in Vespertilionidae. Based on their findings, which indicated major differences between miniopterines and other vespertilionids, Gopalakrishna and Chari recommended removal of Miniopterus to its own family, Miniopteridae. Pierson (1986) proposed a series of phylogenetic hypotheses based on an analysis of transferrin immunological distance data (fig. 13). Her results differed significantly from those of previous studies in many areas, including (1) placing Rhinopomatidae within Rhinolophoidea, (2) associating Mystacinidae with Noctilionoidea rather than Vespertilionoidea, (3) placing Furipteridae and Natalidae at the base of the microchiropteran tree, and (4) grouping Tomopeas and Miniopterus with Molossidae rather than with Vespertilioninae. In the context of Pierson’s trees, Yinochiroptera is monophyletic, but Yangochiroptera and the four superfamilies are not. One of the more striking results of Pierson’s (1986) immunological study—the placement of Mystacinidae—was explored in greater depth in Pierson et al. (1986), which also included a discussion of some morphological features. Hill and Harrison (1987) reviewed craniodental morphology and structure of the baculum in selected vespertilionids, and proposed a new generic classification of those forms traditionally included in Vespertilioninae (e.g., by Miller, 1907). Hill and Harrison (1987) included Myotini and Antrozoini 1998 35SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS Fig. 18. Fenton’s (1992) phylogeny of bat families with Altringham’s (1996) addition of subfamilies (redrawn from Fenton [1992: p. 10] and Altringham [1996: fig. 1.12]). Fenton’s (1992) caption read ‘‘A family tree or phylogeny shows the presumed evolutionary relationships between the living families of bats. The families are grouped according to superfamilies and suborders.’’ Altringham’s (1996) caption read ‘‘An evolutionary tree for the bats to subfamily level. Considerable uncertainty surrounds some parts of this tree, in particular the relationships between megabats and microbats, and the detailed classification of the Phyllostomidae and Vespertilionidae.’’ as tribes within Vespertilioninae, and recognized Nyctophininae as a distinct subfamily. Tiunov (1989) examined variation in morphology of the tongue and male accessory glands (e.g., prostate, seminal vessicles, Cowper’s gland, ampullary glands) in a number of Old World species, and discussed the phylogenetic implications of these data. Based on observed differences, Tiunov concurred with Gopalakrishna and Chari’s (1983) recommendation that Miniopteridae should be recognized as a family distinct from Vespertilionidae. In contrast, Tiunov found no differences between rhinolophines and hipposiderines, and therefore recommended that they be placed in a single family, Rhinolophidae. Baker et al. (1991a) analyzed variation in rDNA restriction sites in a study designed to 36 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY Fig. 19. Volleth and Heller’s hypothesis of relationships among vespertilionids (redrawn from Volleth and Heller, 1984: fig. 7). The original caption read ‘‘Cladogram of Vespertilionidae based on karyological features . . . Where the names of species are lacking, only one species was studied (12 cases). From karyologically heterogeneous genera all species studied are shown. The dotted line between Eptesicus and Scotophilus indicates a second possibility for the relations of Eptescini . . .[that tribe] could be grouped together with Scotophilus ....[Abbreviations:] H. 5 Hesperopternus,Hyps. 5 Hypsugo, Ny. 5 Nyctalus,P. 5 Pipistrellus,T. 5 Tylonycteris.’’ 1998 37SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS test bat monophyly (see below). Although their data set could not resolve subordinal or interfamilial relationships of bats, support was found for several groups, including Yinochiroptera, Noctilionoidea, and a clade containing Mormoopidae and Noctilionidae. In the 1980s and early 1990s, a significant controversy arose in chiropteran systematics regarding bat monophyly. The history of this controversy was reviewed by Simmons (1994) and will not be repeated here. Although bat diphyly has been proposed by several authors (e.g., Smith and Madkour, 1980; Hill and Smith, 1984; Pettigrew, 1986, 1991a, 1991b, 1994, 1995; Pettigrew and Jamieson, 1987; Pettigrew et al., 1989; Rayner, 1991b; Pettigrew and Kirsch, 1995), a growing body of data provides very strong support for bat monophyly. Data supporting chiropteran monophyly include morphological features from many organ systems (Luckett, 1980a, 1993; Wible and Novacek, 1988; Kovtun, 1989; Thewissen and Babcock, 1991, 1993; Kay et al., 1992; Novacek, 1992, 1994; Beard, 1993; Simmons, 1993a, 1994, 1995; Wible and Martin, 1993; Simmons and Quinn, 1994; Miyamoto, 1996), DNA–DNA hybridization data (Kirsch et al., 1995; Hutcheon and Kirsch, 1996; Kirsch, 1996), and sequence data from numerous mitochondrial and nuclear genes (Adkins and Honeycutt, 1991, 1993, 1994; Mindell et al., 1991; Ammerman and Hillis, 1992; Bailey et al., 1992; Stanhope et al., 1992, 1993, 1996; Honeycutt and Adkins, 1993; Knight and Mindell, 1993; Novacek, 1994; Allard et al., 1996; Miyamoto, 1996; Porter et al., 1996). Bat monophyly now represents one of the most strongly supported phylogenetic hypotheses within Mammalia (Simmons, 1994; Miyamoto, 1996). Novacek (1991) used two novel phylogenies as a framework for discussing evolution of cochlear features in bats (fig. 16). These trees were not derived from an explicit character analysis, but were instead based on consideration of the morphological characters discussed in Koopman (1984) and Novacek (1987) as well as Pierson’s (1986) immunological results (see caption to fig. 16). Both of Novacek’s (1991) trees indicated paraphyly of Yinochiroptera and Emballonuroidea, left monophyly of Yangochiroptera unresolved, and suggested monophyly of the remaining three superfamilies. Monophyly of Vespertilionidae (including kerivoulines, tomopeatines, and miniopterines) was not discussed. Griffiths and his colleagues used osteomyological characters of the hyoid region to explore relationships among various groups of yinochiropteran bats (Griffiths and Smith, 1991; Griffiths et al., 1992). Griffiths et al. (1992) presented two alternative phylogenies (fig. 17), neither of which supported monophyly of Emballonuroidea or Rhinolophoidea. The only clade that appeared in both trees was Rhinolophidae 1 Hipposideridae. Gopalakrishna and Badwaik (1992) discussed fetal membrane structure and used phenetic similarity to evaluate phylogenetic relationships among bat families. They concluded that . . . similarities between Molossidae and Pteropodidae and differences between Molossidae and Vespertilionidae suggest a closer relationship between Pteropodidae and Molossidae than between Molossidae and Vespertilionidae. It is, therefore, suggested on purely embryological grounds that Molossidae be separated from the Super-family Vespertilionidae and be placed somewhere between Pteropodidae and Emballonuridae (Gopalakrishna and Badwaik, 1992: 7). Fenton (1992) provided a phylogeny of bat families in his semipopular book on bats, but did not discuss the source of this hypothesis. The same topology was reproduced by Fenton (1995). This tree (fig. 18) was developed to summarize possible relationships of bats as reflected in numerous systematic studies and classifications (e.g., Hill, 1974; Pierson et al., 1986); it was not based on any new data or data analyses (Fenton, personal commun.). Altringham (1996) reproduced Fenton’s tree (again with no mention of source), modifying it only by adding branches for each subfamily (fig. 18). Volleth and Heller (1994) provided the first cladistic hypothesis of relationships among genera of Vespertilionidae sensu lato (fig. 19). Using data from G-banded chromosomes, they identified homologous arms and presented a cladogram based on an analysis of derived chromosomal features (e.g., translocations, Robertsonian fusions, fissions). They concluded that (1) Miniopterinae is the sister-group to a clade containing 38 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY Fig. 20. Results of Simmons’ (1998) analysis of higher-level relationships based on morphology and rDNA restriction sites. The original caption of this figure read ‘‘Strict consensus of two equally most-parsimonious trees (608 steps each; CI 5 0.410, RI 5 0.592) found in a heuristic analysis . . . The numbers below internal branches indicate the percentage of bootstrap replicates in which each clade appeared; numbers above the branches are decay values (the minimum number of additional steps required to collapse each clade).’’ Kerivoulinae 1 Murininae 1 Vespertilioninae; (2) Myotini falls outside a clade containing the remaining vespertilionines; (3) Nyctophilus is a member of Vespertilionini; (4) Vespertilionini and Pipistrellini are sistertaxa; and (5) Eptesicini (including Hesperoptenus), ‘‘Nycticeiini,’’ and Plecotini fall outside the Vespertilionini 1 Pipistrellini clade. Antrozous,Tomopeas,Lasiurus, and other New World taxa were omitted from Volleth and Heller’s study. Sudman et al. (1994), drawing in part upon the unpublished work of Barkley (1984), investigated relationships of Tomopeas (the only member of Tomopeatinae) to vespertilionids and molossids using protein electrophoresis, cytochrome bgene sequences, and morphological characters. They concluded that these data support a close relationship between Tomopeas and molossids rather than vespertilionids, and recommended that Tomopeatinae be transferred from Vespertilionidae to Molossidae. Kirsch and his colleagues (Pettigrew and Kirsch, 1995; Hutcheon and Kirsch, 1996; Kirsch, 1996; Kirsch and Hutcheon, 1997; Hutcheon et al., in press; Kirsch and Pettigrew, in press; Pettigrew and Kirsch, in press) reported results of a series of DNADNA hybridization studies of relationships among a set of taxa including both megachiropterans and microchiropterans. Surprisingly, these experiments suggested that Rhinolophoidea and Pteropodidae may be sistertaxa, implying microchiropteran paraphyly. However, this result has been questioned by 1998 39SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS TABLE 2 Higher-level Classification of Recent Bats Proposed by Simmons (1998) Order Chiroptera Suborder Megachiroptera Family Pteropodidae Suborder Microchiroptera Infraorder incertae sedis Superfamily Emballonuroidea Family Emballonuridae Infraorder Yinochiroptera Superfamily Rhinopomatoidea Family Craseonycteridae Family Rhinopomatidae Superfamily Rhinolophoidea Family Nycteridae Family Megadermatidae Family Rhinolophidae Subfamily Rhinolophinae Subfamily Hipposiderinae Infraorder Yangochiroptera Superfamily incertae sedis Family Mystacinidae Superfamily Noctilionoidea Family Noctilionidae Family Mormoopidae Family Phyllostomidae Superfamily Molossoidea Family Antrozoidae Family Molossidae Subfamily Tomopeatinae Subfamily Molossinae Superfamily Vespertilionoidea Family Vespertilionidae Subfamily Vespertilioninae Subfamily Miniopterinae Subfamily Myotinae Subfamily Murininae Subfamily Kerivoulinae Superfamily Nataloidea Family Myzopodidae Family Thyropteridae Family Furipteridae Family Natalidae the authors themselves, who have suggested that base compositional bias might be responsible for producing a false phylogenetic signal linking two A-T rich clades (Pteropodidae and Rhinolophoidea; Hutcheon and Kirsch, 1996; Kirsch, 1996; Kirsch and Hutcheon, 1997; Hutcheon et al., in review; Kirsch and Pettigrew, in review; Pettigrew and Kirsch, in review). Although topology of other parts of the tree remains suspect, many recovered groupings are congruent with those identified in previous studies. As noted by Hutcheon et al. (in press), ‘‘a tree as startling as ours obviously must be verified by additional studies . . .’’ The most recent comprehensive attempt to resolve higher-level relationships among extant family-level lineages of bats was that of Simmons (1998). She assembled a data set of 192 discrete characters including many new morphological characters, all of the relevant morphological data discussed by most previous workers (e.g., Van Valen, 1979; Luckett, 1980a; Novacek, 1980a, 1991; Barkley, 1984; Griffiths and Smith, 1991; Griffiths et al., 1992), and the rDNA restriction site data presented by Baker et al. (1991a). To test vespertilionid monophyly and evaluate relationships of its subfamilies, Vespertilionidae sensu lato was split into seven subgroups for analysis: Vespertilioninae, Myotinae, Miniopterinae, Murininae, Kerivoulinae, Antrozoinae, and Tomopeatinae. Parsimony analyses of the resulting data set produced a well-resolved tree (fig. 20) in which many nodes were strongly supported. Major results included the following: (1) Emballonuridae appears to be the sistergroup of all other microchiropterans, therefore Emballonuroidea and Yinochiroptera (as traditionally recognized) are not monophyletic; (2) Rhinopomatidae and Craseonycteridae are sister-taxa; (3) Rhinolophoidea, Noctilionoidea, Vespertilionoidea, and Yangochiroptera each appear to be monophyletic; (4) Vespertilionidae sensu lato is not monophyletic; Antrozoinae and Tomopeatinae are more closely related to Molossidae than to other vespertilionids; and (5) Myzopodidae, Thyropteridae, Natalidae, and Furipteridae form a clade. Simmons (1998) proposed a number of nomenclatural changes (table 2) based on her phylogenetic results, restricting Emballonuroidea to Emballonuridae, placing Rhinopomatidae and Craseonycteridae in Rhinopomatoidea, restricting Yinochiroptera to Rhinopomatoidea 1 Rhinolophoidea, raising Antrozoinae to family rank as Antrozoidae, referring Tomopeatinae to Molossidae, recognizing Antrozoidae 1 Molossidae as a new superfamily Molossoidea, recognizing Myzopodidae 1 Thyropteridae 1 Natalidae 40 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY 1 Furipteridae as a new superfamily Nataloidea, and restricting Vespertilionidae to Vespertilioninae 1 Miniopterinae 1 Myotinae 1 Murininae 1 Kerivoulinae. Noting that several of the latter subfamilies might eventually be raised to family rank, Simmons (1998) restricted Vespertilionoidea to Vespertilionidae as defined above, pending further study. GOALS OF THE CURRENT STUDY The data set developed by Simmons (1998)—which of course includes many characters originally described by others (see appendix 2)—includes more than 80 craniodental and postcranial osteological characters. This provides a unique opportunity to evaluate the evolutionary relationships of the archaic Eocene taxa in the context of a phylogeny of extant lineages, and to investigate methodological issues involving fossil taxa and missing data in phylogenetic analyses. The goals of the current study are as follows: (1) to determine the relationships of Icaronycteris,Archaeonycteris,Hassianycteris, and Palaeochiropteryx to each other and to family-level lineages of extant bats; (2) to evaluate the effects of including fossil genera on perceived relationships among extant forms; (3) to evaluate the effects of including softtissue and molecular characters in a study including both fossil and extant groups; (4) to document patterns of character transformation in the basal part of the chiropteran tree; and (5) to consider the implications of these data for theories concerning the early evolution of echolocation and foraging strategies in Microchiroptera. MATERIALS AND METHODS TAXONOMIC SAMPLING AND MONOPHYLY The present study consists of a phylogenetic analysis of 24 family-level lineages of extant bats, two ordinal-level extant outgroup taxa (Scandentia, Dermoptera), and four Eocene fossil bat genera (Icaronycteris,Archaeonycteris,Palaeochiropteryx,Hassianycteris; see appendix 1 for specimens examined). The extant lineages are the same as those analyzed by Simmons (1998); taxonomic names used for bat clades follow the classification proposed by Simmons (1998; table 2). Monophyly of Vespertilioninae (including Nyctophilinae) was assumed for practical reasons following Volleth and Heller (1994) and Simmons (1998). Monophyly of each of the other extant bat lineages used as OTUs (operational taxonomic units) is well established (see Simmons [1998] and references cited therein). Each of the fossil genera also appears to be monophyletic (table 3), with the possible exception of Icaronycteris. In the absence of apomorphic traits, most workers have diagnosed Icaronycteris on the basis of primitive features that have been modified in all other bats (Jepsen, 1966, 1970; Novacek, 1987). Although Jepsen (1966, 1970) noted 14 characters of Icaronycteris that he considered as ‘‘primitive’’ or ‘‘generalized’’ relative to the conditions seen in extant bats, close examination of these features by Novacek (1987) reduced this list considerably. Novacek identified only four primitive features apparently found in Icaronycteris that have been lost or modified in all other known fossil and extant bats: (1) unfused sternal elements, (2) relatively short radius, (3) complete phalangeal formula (23-3-3) on the digits of the wings, and (4) head and neck of femur set at an angle to the shaft. Our examination of additional specimens of extant bats plus material of Icaronycteris that was not available to Jepsen or Novacek indicate that two more of these features should be removed from the list. Our observations indicate that sternal elements are typically unfused in juvenile, subadult, and some young adult bats. Novacek (1987) noted that the holotype of Icaronycteris index (PU 18150) might have unfused sternal elements simply because it was a young individual. This was confirmed by our examination of UW 21481a–b (fig. 21), a previously undescribed specimen of Icaronycteris index from Wyoming. 2 The sternal el2 We identified UW 21481a–b as Icaronycteris index 1998 41SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS ements in this individual are fused with the exception of the joints between the manubrium and mesosternum and between the mesosternum and xiphisternum. Sutures are visible between three mesosternal elements, but the remaining mesosternal elements are fully fused (fig. 22). This degree of fusion is similar to that found in young adult bats of many extant families. Because UW 21481a–b and PU 18150 are similar in all other respects, we conclude that PU 18150 was a young adult Icaronycteris, whereas UW 21481a–b represents a somewhat older individual. A similar pattern of variation is also seen in some of the Messel bats. For example, all mesosternal elements are fused in some specimens of Archaeonycteris trigonodon (e.g., SMF Me 80/1379), while all sutures are still visible in others (e.g., SMF Me 963a). Icaronycteris therefore appears to have had an ontogenetic pattern of sternal fusion similar to that seen in Archaeonycteris and many extant bats. based on morphology and collection locality. Based on comparisons with Jepsens original stereophotographs (many of which remain unpublished), we found that UW 21481a–b appears identical to the holotype of Icaronycteris index PU 18150 in vitually all respects (although see text for a discussion of sternal fusion). Forearm length in UW 21481a–b is 47.5 mm; forearm length in PU 18150 is 48.0 mm. Measurements of skull length, tibia length, and length of various wing elements in UW 21481a–b are within 6 2 mm of those reported by Jepsen (1966) for PU 18150. This range of variation is well within that seen in extant species of similar body size (e.g., Swanepoel and Genoways, 1979). UW 21481a–b was collected from early Eocene sediments of the Green River Formation of Wyoming, the same deposits from which PU 18150 was collected (Jepsen, 1966). Because detailed dental comparisons between PU 18150 and UW 21481a–b were precluded by preservation of the latter specimen, it remains possible that UW 21481a–b represents a species of Icaronycteris distinct from index; however, we found no data to support such an interpretation. Accordingly, we refer UW 21481a–b to Icaronycteris index Jepsen, 1966. Grande (1980: fig. III.26) figured a badly preserved bat fossil from Middle Eocene deposits of the Green River Formation of Colorado that he noted ‘‘probably represents an undescribed species.’’ Unfortunately, this specimen is in a private collection and has never been described. This record suggests, however, that bat species other than Icaronycteris index may have been present in western North America in the Eocene. Another feature cited by Jepsen (1966, 1970) and Novacek (1987) as a primitive character of Icaronycteris is the orientation of the head and neck of femur, which they described as set at an angle to the long axis of the shaft (although not to the extent seen in terrestrial mammals). Novacek (1987: 13) noted that in this trait Icaronycteris is ‘‘clearly more conservative than living megaand microchiropterans, where the head is nearly aligned with the shaft and the neck is very short or absent.’’ However, our comparisons of the femur of Icaronycteris with those of extant bats indicate that Novacek was misled by the angle of presentation of the femur in PU 18150, which provides an oblique view of the head and neck (fig. 23). We found that this view accentuates a relatively small degree of offset of the femoral head. Comparisons of the femur of PU 18150 and UW 21481a–b with those of extant megachiropterans and noctilionoids of similar body size indicate that the head and neck of the femur of Icaronycteris are set at the same angle (relative to the shaft of the femur) as seen in many extant bats. Indeed, the proximal femur of Rousettus aegyptiacus is virtually identical to that of Icaronycteris when viewed from the same angle. Icaronycteris thus cannot be considered more primitive than extant bats in terms of femur morphology. Our survey thus limits the list of primitive features found in Icaronycteris (but modified in all other known bats) to the following: (1) relatively short radius, and (2) complete phalangeal formula (2-3-3-3) on the digits of the wings. Novacek (1987) noted that relative length of the radius had never been quantified, but this was subsequently done by Habersetzer and Storch (1987). They found that the ratio of humerus length to radius length was somewhat larger (and the radius therefore relatively shorter) in Icaronycteris compared to other Eocene bats, although the difference between the range of values for Icaronycteris (0.715–0.732) and Archaeonycteris (0.680–0.715) is not great. The diagnosis of Icaronycteris therefore rests on just two plesiomorphic features. However, it seems unlikely that multiple lineages are represented in this group (at least among the four specimens that we examined) because the range of variation in size and 42 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY TABLE 3 Apomorphies a Diagnosing Genera of Eocene Bats Icaronycteris (one species, I. index) ● No apomorphies Archaeonycteris (three species, A. trigonodon, A. pollex, A. brailloni) ● Ectoflexus deeply retracted between parastyle and metastyle on M1 and M2 ● Ventral process of manubrium of sternum oriented at approximately 90 8 to axis of body of manubium ● Posterior xiphisternum narrow, without lateral flare ● Dorsal ischial tuberosity present Palaeochiropteryx (two species, P. tupaiodon and P. spiegeli) ● Metaconule present on M1 and M2 ● Length of postparacrista equals length of premetacrista on M3 ● Ventral accessory process present on cervical vertebra 5 ● Infraspinous fossa of scapula relatively broad ● Ventral projection present on anteromedial flange of scapula Hassianycteris (four species, H. messelensis, H. magna, H. revilliodi, and H. joeli) ● Protoconid and hypoconid on molar teeth exceptionally tall and robust ● Last lower premolar not molariform, metaconid lacking and talonid short ● First upper premolar reduced to tiny peg or absent ● Mandible deep, thickened dorsoventrally ● Radius unusually long and strongly curved ● Trochiter extends well beyond level of head humerus ● Metacarpal of digit V relatively short compared to metacarpals III and IV a Not found in any other Eocene taxon. Membership in extant families is precluded by documented or inferred absence of derived characters diagnostic of those taxa (for a list of characters diagnosing extant lineages see Simmons, 1998: tables 1, 2). Sources: Russell and Sige´, 1970; Smith and Storch, 1981; Habersetzer and Storch, 1987, 1989; Storch and Habersetzer, 1988; personal obs.). shape of skeletal elements among referred individuals appears to fall within limits characteristic of extant species (see footnote 2). We have therefore assumed monophyly of Icaronycteris for the purposes of the current study, although we recognize that the European material referred to ?Icaronycteris menui and ?Icaronycteris sp. may represent one or more different clades. OUTGROUPS Only one outgroup is necessary to root a phylogenetic tree (Nixon and Carpenter, 1993), but at least two outgroups are usually included in cladistic analyses to establish character polarities and to permit testing of ingroup monophyly (Maddison et al., 1984). Ideally, outgroups should comprise the nearest sister-taxa to the ingroup because the probability of homoplasy increases with time since divergence from a common ancestor. The identity of the sister-group of bats is still the subject of considerable debate, and recent congruence studies have failed to resolve existing conflicts between morphological data and several molecular data sets (e.g., Allard et al., 1996; Miyamoto, 1996; Stanhope et al., 1996). Nevertheless, practical considerations require that a choice of outgroups be made to facilitate analyses of relationships and character evolution within Chiroptera. A close relationship between Chiroptera and Dermoptera (colugos or flying lemurs) is strongly supported by morphological data (Wible and Novacek, 1988; Novacek, 1992, 1994; Simmons, 1993a, 1994, 1995; Szalay and Lucas, 1993; Wible, 1993; Simmons and Quinn, 1994; Miyamoto, 1996; Griffiths, MS), 12S rDNA sequences (sampled from more than 150 species representing 20 orders; Vrana, 1994), and combined morphological data and cytochrome oxidase II (COII) gene sequences (Novacek, 1994). Accordingly, Simmons (1998) used Dermoptera as an outgroup for investigating interrelationships of bats. Morphological synapomor- 1998 49SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS maxillae that lie on the face adjacent to the external narial opening. The nasal branches of the premaxillae are relatively well developed in Icaronycteris,Archaeonycteris,Hassianycteris, and Palaeochiropteryx, as they are in most extant bats. In contrast, the nasal branches of the premaxillae in rhinolophoids are either entirely absent or are reduced to tiny splints of bone. The nasal branches of the premaxillae are well developed in both outgroups, suggesting that the state seen in the Eocene fossil bats is primitive. Character 11: Palatal branches of premaxillae well developed (0); or reduced or absent (1). The palatal branches of the premaxillae are defined as those portions of the premaxillae that contribute to the anterior palate. The palatal branches of the premaxillae are relatively well developed in Icaronycteris and many extant bats (Nycteridae, Rhinolophidae, Phyllostomidae, Noctilionidae, Mormoopidae, Mystacinidae, Myzopodidae, Thyropteridae, Natalidae, some Molossinae). In contrast, the palatal branches of the premaxillae are either reduced to tiny splints of bone or are entirely absent in other extant lineages. This reduction is apparently a derived condition since well developed palatal branches are present in both outgroups and most other mammals. The condition of the anterior palate in Archaeonycteris and Palaeochiropteryx could not be determined from available specimens. We found that the palatal branches of the premaxillaries are not visible in specimens prepared in dorsal view, and that the mandibular rami obscure the palate in all specimens prepared in ventral view. Accordingly, these forms are scored ‘‘?’’ for this character. Smith and Storch (1981: 154, 164) noted that in Hassianycteris the ‘‘palatal branch [of the premaxilla] not well developed, premaxillaries not fused [at midline],’’ and that ‘‘the apparent shape and reduction of the premaxillary is quite reminiscent of the derived condition of this element in emballonurids.’’ However, the source(s) of these observations were not discussed, and we were unable to confirm them during our microscopic examination of specimens. Smith and Storch apparently described these features based on radiographs, although the published versions are too fuzzy to be of use. In both published radiographs (Smith and Storch, 1981: figs. 1, 2), an image of the anterior dentition is superimposed on the anterior palate. Considering the complex nature of this region and the high density of the dental images, any observations of palatal morphology in these specimens must be viewed with caution. We considered scoring Hassianycteris as having a reduced palatal branch of the premaxilla based on the report by Smith and Storch (1981), but subsequently noted that reduction and lack of fusion of the palatal branches were omitted from Habersetzer and Storch’s (1987) diagnosis of Hassianycterididae (which contains only Hassianycteris). Given the prominence of these features in Smith and Storch’s original diagnosis of Hassianycteris, we can only conclude that there is considerable uncertainty about the structure of the premaxilla in these bats. Accordingly, we score Hassianycteris as ‘‘?’’ for this character. Character 15: Hard palate extends posteriorly into interorbital region (0); or terminates either at or anterior to level of zygomatic roots (1). The hard palate, which forms a bony separation between the oral and nasal passages, terminates posteriorly at the mesopterygoid fossae. Position of the posterior edge of the hard palate varies independently of the posterior extent of the molar toothrow, and is apparently linked to the structure of the nasal passages and soft tissues of the pharyngeal region. The hard palate extends posteriorly into the interorbital region in Archaeonycteris,Palaeochiropteryx, one of the outgroups (Scandentia), and many extant bats (Pteropodidae, some Emballonuridae, some Hipposiderinae, some Phyllostomidae, Mormoopidae, Noctilionidae, Mystacinidae, Myzopodidae, Thyropteridae, some Furipteridae, Natalidae, Antrozoidae, Tomopeatinae, some Molossidae, and Vespertilionidae). In contrast, the hard palate terminates either at or anterior to the level of the zygomatic roots in the other outgroup (Dermoptera) and all remaining extant bats. Lack of agreement between the two outgroups precludes a priori determination of the primitive condition for this character. The postition of the posterior end of the hard palate in Icaronycteris and Hassianycteris could not be determined due to the position of the 50 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY mandibular rami in all avialable specimens, so these forms are scored ‘‘?’’ for this character. Character 16: Two upper incisors in each side of jaw (0); or one incisor (1); or upper incisors absent (2). Because of difficulties associated with determining homologies of the anterior dental loci among different bat lineages (reviewed by Slaughter, 1970), we chose to score the number of teeth in various dental regions (i.e., upper incisors, lower incisors, upper premolars, lower premolars) rather than attempting to score presence/absence of teeth at specific loci. Although some potentially informative patterns may be overlooked by this method, this approach preserves that homology information of which we are most confident and permits scoring of all taxa (including the outgroups) for each character. The size of one or more teeth in a number of species has been reduced to the point where these teeth are considered vestigial (Slaughter, 1970). This is particularly common in the premolar dentition. In some instances there is within-species (and even within-individual) polymorphism with respect to presence/absence of vestigial teeth (e.g., the anterior upper premolar in Rhinolophus clivosus, the middle lower premolar in Chrotopterus auritus). In our experience, older individuals frequently lack such teeth, while they are often present in younger animals. This pattern suggests that vestigial teeth are often lost during the lifetime of the individual. Consequently, we have scored the taxa in question as having the higher dental formula in cases of within-species polymorphism in presence/absence of vestigial teeth. Two upper incisors are present on each side of the jaw in Archaeonycteris,Hassianycteris, and Palaeochiropteryx. This is the same upper incisor formula as seen in many extant bats (some Pteropodidae, some Emballonuridae, Nycteridae, some Phyllostomidae, Mormoopidae, Noctilionidae, Nataloidea, some Vespertilioninae, Miniopterinae, Myotinae, Murininae, and Kerivoulinae). In contrast, only one incisor is present on each side of the upper jaw in some Pteropodidae, some Emballonuridae, Rhinopomatoidea, Rhinolophidae, some Phyllostomidae, Mystacinidae, Molossoidea, and some Vespertilioninae. The upper incisors are entirely absent in some Pteropodidae and Megadermatidae. Both outgroups have two upper incisors on each side of the jaw. In this context, the state seen in Archaeonycteris,Hassianycteris, and Palaeochiropteryx apparently represents the primitive condition. We were unable to unambiguously determine the number of upper incisors in Icaronycteris, so this taxon is scored ‘‘?’’ for this character. Character 17: Three lower incisors in each side of jaw (0); or two incisors (1); or one incisor (2); or incisors absent (3). Three lower incisors are present on each side of the jaw in Icaronycteris,Archaeonycteris,Hassianycteris, and Palaeochiropteryx. This is the same lower incisor formula seen in many extant bats (some Emballonuridae, Nycteridae, Nataloidea, some Antrozoidae, some Molossinae, and Vespertilionidae). Two lower incisors are present on each side of the lower jaw in some Pteropodidae, some Emballonuridae, Rhinopomatoidea, Megadermatidae, Rhinolophidae, some Phyllostomidae, Mormoopidae, some Antrozoidae, Tomopeatinae, and some Molossinae. A single lower incisor is present on each side in some Pteropodidae, some Phyllostomidae, Noctilionidae, Mystacinidae, and some Molossinae. Finally, the lower incisors are entirely absent in some Pteropodidae and some Phyllostomidae. Both outgroups have three lower incisors on each side. In this context, the state seen in the Eocene bat genera apparently represents the primitive condition. Character 18: Three upper premolars in each side of jaw (0); or two premolars (1); or one premolar (2). Three upper premolars are present on each side of the jaw in Icaronycteris,Archaeonycteris,Hassianycteris, and Palaeochiropteryx magna.Hassianycteris messelensis has only two upper premolars (fig. 24), apparently having lost the tiny peglike anterior premolar seen in H.magna. Among extant bats, three upper premolars are present in some Pteropodidae, some Phyllostomidae, Myzopodidae, Thyropteridae, Natalidae, some Myotinae, and Kerivoulinae. In contrast, two upper premolars are present on each side of the jaw in some Pteropodidae, Emballonuridae, some Megadermatidae, Rhinolophinae, some Hipposiderinae, some Phyllostomidae, Mormoopidae, 1998 51SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS Fig. 24. Hassianycteris messelensis (SMF ME 1414a). Closeup lateral view of skull. Note the robust dentary and dentition, and the absence of the first upper premolar. A vestigal first upper premolar is present in H.magna. AN, angular process of dentary; VAP, ventral accessory process of vertebra (C3?). Scale bar 5 10 mm. From Habersetzer and Storch (1987: fig. 5). Photo by E. Pantak (Senkenbergmuseum). Mystacinidae, Furipteridae, some Molossinae, some Vespertilioninae, Miniopterinae, some Myotinae, and Murininae. A single upper premolar is present on each side in Rhinopomatoidea, Nycteridae, some Megadermatidae, some Hipposiderinae, some Phyllostomidae, Noctilionidae, Antrozoidae, Tomopeatinae, some Molossinae, and some Vespertilioninae. Among the outgroups, Scandentia has three upper premolars, while Dermoptera has two. Lack of agreement between the two outgroups precludes a priori determination of the primitive condition for this character. Character 19:Middle upper premolar with three roots (0); or with two roots (1); or with one root (2). Among bats with three upper premolars, the middle premolar (usually designated P3) may have either one, two, or three roots. P3 has three roots in Icaronycteris. In contrast, this tooth has only two roots in Archaeonycteris,Hassianycteris, and Palaeochiropteryx. Among extant bats, P3 is triple-rooted only in Natalidae. P3 has two roots in Pteropodidae, some Phyllostomidae, Thyropteridae, and some Kerivoulinae. P3 is single-rooted in some Phyllostomidae, Myzopodidae, Myotinae, and some Kerivoulinae. Three roots are present on P3 in both outgroups, suggesting that the triple-rooted condition seen in Icaronycteris is primitive. The double-rooted condition seen in the other Eocene taxa is apparently derived. Because it is thought that P3 is often the first tooth lost from the premolar dentition in species with a reduced number of teeth (Slaughter, 1970), we scored this character only in forms that retain three upper premolars (state ‘‘0’’ of character 18 above). Accordingly, Emballonuridae, Yinochiroptera, Mormoopidae, Noctilionidae, Mystacinidae, Furipteridae, Molossoidea, Vespertilioninae, Miniopterinae, and Murininae are scored ‘‘-’’ for this character. Character 20: Three lower premolars in each side of jaw (0); or two premolars (1). Three premolars are present on each side of the lower jaw in Icaronycteris,Archaeonycteris,Hassianycteris and Palaeochiropteryx. Among extant bats, three lower premolars are present in Rhinolophinae, some Phyllostomidae, Mormoopidae, Nataloidea, some Vespertilioninae, Miniopterinae, some Myotinae, and Kerivoulinae. Only two lower premolars are present in all other lineages. Among the outgroups, Scandentia has three lower premolars, while Dermoptera has two. Lack of agreement between the two outgroups precludes a priori determination of the primitive condition for this character. Character 21: Lower first and second molars with primitive tribosphenic arrangement of cusps and cristids (0); or nyctalodont (1); or myotodont (2); or teeth modified for fruit and/or nectar or blood feeding, cusps and cristids not distinct (3). Menu and Sige´ (1971) discussed molar morphology in bats, and distinguished two forms of the talonid of m1 and m2 that they termed ‘‘nyctalodonty’’ and ‘‘myotodonty.’’ In both morphotypes, the hypoconulid lies adjacent to the entoconid on the lingual edge of the tooth, having shifted lingually from the midline position (where it is more or less equidistant from the labial and lingual borders of the tooth) characteristic of primitive therian tribosphenic molars. Nyctalodonty is defined by the position of the postcristid, which connects the hypoconid with the hypoconulid in nyctalodont forms. Myotodonty is distinguished by an alternative arrangement in which the post- 52 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY cristid bypasses the hypoconulid to connect instead with the entoconid. The postcristid is relatively high and forms a sharp shearing edge in both conditions, but no cristid runs between the entoconid and hypoconulid. Although it is easy to classify molar morphology of most extant bats, Menu and Sige´ (1971) noted that problems arise when Eocene bats are considered because the hypoconulid is located in the midline position in some taxa, always connected to the hypoconid and sometimes connected to the entoconid by low cristids. The posterior edge of the talonid is thus dominated by a large conical cusp (the hypoconulid) rather than by a shearing crest (the postcristid). This morphology, termed ‘‘archaic’’ by Menu and Sige´ (1971), appears in numerous Late Cretaceous and Paleocene mammals, including forms thought to represent the earliest members of several eutherian radiations (e.g., primates, insectivores; Menu and Sige´, 1971). This apparently represents the primitive tribosphenic condition. Possible differences in function of the three talonid types discussed above have not yet been examined. Nyctalodont and myotodont teeth are at least superficially similar in terms of the length and height of the postcristid, which may indicate an increased reliance on shearing as compared to the primitive tribosphenic condition. Increased shearing has been associated with specializations for feeding on hard-shelled beetles (Freeman, 1979) and on soft-bodied prey (Strait, 1993). Freeman (1979) found that molossid bats that feed on beetles have relatively larger teeth with longer shearing cusps than do molossids that feed on soft-bodied prey. Alternatively, Strait (1993) found that mammalian insectivores that feed predominantly on soft-bodied prey (e.g., moths and caterpillars) have relatively longer shearing crests (summed across the entire molar) than do close relatives that have more generalized feeding habits and consume large quantities of hard-bodied beetles. Whatever the relationship between shearing and prey type, these studies suggest that variation in talonid form may be indicative of differences in dietary habits. However, there does not seem to be any obvious correlation between feeding habits and talonid type in extant bats. For example, rhinolophids that feed on soft-bodied prey (e.g., Rhinolophus blasii) and those that feed on hard-bodied prey (e.g., Hipposideros commersoni) both have nyctalodont teeth (Strait, 1993). It therefore seems premature to draw any functional conclusions from transformations in talonid structure in bats. The first lower molars exhibit the primitive tribosphenic arrangement of the talonid cusps and cristids in Archaeonycteris. In contrast, the lower molars are nyctalodont in Hassianycteris. Both nyctalodont and myotodont conditions appear in Palaeochiropteryx, although never in the same individual. Both nyctalodont and myotodont specimens have been referred to Palaeochiropteryx tupaiodon (e.g., by Russel and Sige´, 1970), raising the possiblity of within-species polymorphism. However, we are unaware of any cases of within-species polymorphism in this character in extant bats. This suggests that P. tupaiodon may be a composite species, a hypothesis that will require testing with additional data. Meanwhile, we score Palaeochiropteryx as exhibiting both nyctalodonty and myotodonty. Van Valen (1979) claimed that the lower molars of Icaronycteris are nyctalodont, an observation that was subsequently cited by other authors (e.g., Gingerich, 1987). We disagree with Van Valen’s assessment. Our observations indicate that Icaronycteris index had primitive tribosphenic lower molars. The European specimens referred to ?Icaronycteris by Russell et al. (1973) include both nyctalodont and primitive tribosphenic morphotypes. However, this material consists of isolated teeth only, and we consider the affinites of these specimens to each other and to Icaronycteris index to be uncertain at best. Accordingly, we score Icaronycteris as exhibiting only the primitive tribosphenic condition. Among the extant families, the primitive tribosphenic condition of the talonid is seen in fossil members of Hipposiderinae (i.e., Palaeophyllophora) and Megadermatidae (i.e., Necromantis; Beard et al., 1992). Nyctalodonty occurs in Emballonuridae, Rhinopomatoidea, Nycteridae, extant Megadermatidae, Rhinolophinae, extant Hipposiderinae, some Phyllostomidae, Mormoopidae, Furipteridae, Natalidae, some Molossinae, some 1998 53SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS Fig. 25. Icaronycteris index (UW 2244). Ventral view of the basicranium and posterior lower jaws. From Novacek (1987: fig. 2). An, angular process of dentary; Co, cochlea; Ect, ectotympanic; Gl, glenoid fossa; Oa, orbicular apophysis; Sl, stylohyal; Z, zygomatic arch. Vespertilioninae, Miniopterinae, some Myotinae, and Murininae. In contrast, myotodonty is found in some Phyllostomidae, Noctilionidae, Mystacinidae, Myzopodidae, Thyropteridae, Antrozoidae, Tomopeatinae, some Molossinae, some Vespertilioninae, some Myotinae, and Kerivoulinae. In Pteropodidae and some Phyllostomidae the teeth are modified for fruit and/or nectar or blood feeding and the cusps and cristids are not distinct. Among the outgroups, Scandentia exhibits the primitive tribosphenic condition, but we could not score Dermoptera owing to its unique dental morphology. Although polarity cannot be ascertained unambiguously, the data suggest that the primitive tribosphenic condition is primitive for bats. Character 22: Lower jaw with elongate angular process (0); or without elongate angular process (1). The angular region of the lower jaw in mammals is the site of insertion of jaw adductor muscles including m. masseter pars superficialis, m. masseter pars profundus, m. zygomaticomandibularis, m. mandibuloauricularis, and m. pterygoideus internus ( 5 medial pterygoid; Turnbull, 1970). In many ‘‘generalized’’ eutherian mammals and some specialized carnivores, a bony extension of the angle—termed the angular process—lies between the insertions of m. pterygoideus internus and m. masseter. These muscles insert both on the angular process and on a common raphe dorsal, posterior, and ventral to the angular process (Turnbull, 1970). This pattern of muscle insertion is seen in bats that have an elongate angular process (Storch, 1968; Kallen and Gans, 1972). Both m. masseter and m. pterygoideus internus act to close and protrude the lower jaw. M. masseter also pulls the ipsilateral dentary (the dentary on the same side of the skull as the muscle) laterally, while m. pterygoideus internus pulls the ipsilateral dentary lingually (Storch, 1968; Kallen and Gans, 1972). The lower jaw has an elongate angular process in Icaronycteris (figs. 25, 26), Archaeonycteris,Hassianycteris (fig. 24), and Palaeochiropteryx (fig. 27). This is similar to the condition seen in most bats. In contrast, the angular process is either very short (relative to its dorsoventral width) or is effectively absent in Pteropodidae, Craseonycteridae, and some Nycteridae. An elongate angular process is present in Scandentia but not in Dermoptera. Lack of agreement between the two outgroups precludes a priori determination of the primitive condition for this character. However, occurrence of an elongate angular process in many presumably primitive eutherians (e.g., leptictids, Asioryctes, lipotyphlans) suggests that presence of an elongate angular process may be primitive for Eutheria. 54 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY Fig. 26. Icaronycteris index (UW 21481a–b). Stereophotographs of a ventral view of the basicranial region (top) with labelled key photograph (bottom center). AN, angular process of dentary, BS, basisphenoid; CP, coronoid process of dentary; CTP, caudal process of tympanic; ER, epitympanic recess; PGF, postglenoid foramen; PR, periotic; SL, tylohyal; SQ, squamosal; Z, zygomatic arch. Photographs by C. Tarka. Character 23: Angular process projects at or below level of occlusal plane of toothrow, well below coronoid process (0); or angular process projects above level of occlusal plane of toothrow, at same level as the coronoid process (1). The angular process, when present, extends from the posteroventral ‘‘corner’’ of the lower jaw. The angular process projects at or below the level of the occlusal plane of the toothrow (well below the coronoid process) in Icaronycteris (figs. 25, 26), Archaeonycteris,Hassianycteris (fig. 24), and Palaeochiropteryx (fig. 27). This is similar to the condition seen in most extant bats and both outgroups. In contrast, the angular process projects above the level of the occlusal plane of the toothrow (roughly at the same level as the coronoid process) in Myzopodidae, Thyropteridae, and Furipteridae. In this context, the condition in the Eocene fossil bats appears to be primitive. This character cannot be evaluated in taxa that lack an angular process (e.g., taxa scored ‘‘1’’ for 22); these forms are therefore scored ‘‘-’’ for this character). 1998 55SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS Fig. 27. Palaeochiropteryx tupaiodon, positive prints of radiographs; from Novacek (1987: fig. 4). A. Dorsoventral view of the skull of SMF ME 788b. B. Lateral view of SMF ME 1127. An, angular process of dentary; Co, cochlea; Cp, coronoid process of dentary; Ect, ectotympanic; Sl, stylohyal; Z, zygomatic arch. Character 24: Postorbital process present (0); or absent (1). The postorbital process is a laterally projecting process of the frontal that forms part of the posterodorsal rim of the orbit. A postorbital process is present in Archaeonycteris,Hassianycteris, and Palaeochiropteryx. Among extant forms, a postorbital process is present only in Pteropodidae, Emballonuridae, Nycteridae, and Megadermatidae. All other extant bats lack a postorbital process. A postorbital process is present in both outgroups, suggesting that presence of this structure represents the primitive condition. We were not able to ascertain whether this process is present in Icaronycteris; this taxon is therefore scored ‘‘?’’ for this feature. Character 25: Pars cochlearis of petrosal sutured to basisphenoid (0); or loosely attached to basisphenoid via ligaments and/or thin splints of bone (1). The petrosal of all extant microchiropterans (with the exception of some Emballonuridae) is not fused with or sutured to the basisphenoid. Instead, the pars cochlearis is loosely attached to the basisphenoid via ligaments and/or thin splints of bone (fig. 28). This condition (sometimes referred to as ‘‘cochlear isolation’’) is thought to function in reducing bone conduction of laryngeal vibrations (Henson, 1970; Van Valen, 1979). Although there is some variation in the degree of isolation of the cochlea as judged by the relative sizes of the openings that surround the periotic (the anterolateral pyriform fenestra, anteromedial to medial parts of the basicochlear fissure, and posteromedial jugular foramen; Novacek, 1991), this ‘‘loosely attached’’ condition clearly differs from the typical mammalian pattern in which the periotic is firmly sutured to the surrounding bones of the basicranium including the basisphenoid. 56 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY Fig. 28. Nyctalus noctula (Vespertilioninae; redrawn from Henson, 1970: fig. 9). Ventral view of the basicranium and audtory region showing major osteological features, blood vessels, and muscles. The ectotympanic and malleus have been reflected laterally on the right side of the skull. APP, anterior process of petrosal; BF, basicochlear fissure; BO, basioccipital; BOP, basioccipital pit; CO, cochlea; ECT, ectotympanic; ETR, epitympanic recess; FC, fenestra cochleae ( 5 round window); FM, foramen magnum; FO, foramen ovale; FV, fenestra vestibuli ( 5 oval window); GF, glenoid fossa; I, incus; ICA, internal carotid artery; M, malleus; MS, m. stapedius; MTT, m. tensor tympani; OAM, orbicular apophysis of malleus; OC, occipital condyle; PF, pyriform fenestra; PGF, postglenoid foramen; PP, paroccipital process; PT, pterygoid hamulus; SF, stapedial fossa; ST, stapes; STA, stapedial artery; T, tendon of m. tensor tympani; TM, tympanic membrane; Z, zygomatic arch. Based on observed gaps and patterns of breakage in the basicranium of preserved specimens of Archaeonycteris and Palaeochiropteryx, it appears that the periotic was loosely attached to the basisphenoid (and the cochlea relatively isolated) in these taxa. In contrast, the periotic is sutured to the basisphenoid in Pteropodidae and some Emballonuridae. A similar condition is seen in both outgroups and most other mammals, suggesting that the ‘‘sutured’’ condition is primitive, and that the ‘‘loosely attached’’ condition seen in Archaeonycteris and Palaeochiropteryx is derived. We were unable to determine the type of periotic connection in Icaronycteris and Hassianycteris; these taxa are therfore scored ‘‘?’’ for this character. Character 26: Cochlea not enlarged (0); or moderately enlarged (1); or greatly enlarged (2). Much has been written about cochlear size and echolocation in bats, with most authors agreeing that cochlear enlargement (specifically enlargement of the basal turn) is a specialization for perceiving the echoes of high-frequency echolocation signals (Henson, 1970; Segall, 1971; Bruns et al., 1983–1984; Novacek, 1985a, 1987, 1991; Habersetzer and Storch, 1992). By plotting maximum external cochlear width against skull length, Novacek (1985a, 1987, 1991) found that extant Microchiroptera and Megachiroptera have nonoverlapping distributions, with microchiropterans having consistently larger cochleae. The megachiropteran condition (i.e., a relatively small cochlea) was interpreted as the primitive condition because it also occurs in many other small to medium-sized mammals (Novacek, 1985a, 1987, 1991). Novacek (1985a, 1987, 1991) also estimated cochlear width and skull length of Icaronycteris and Palaeochiropteryx, and found that these fossil taxa fell within the range of variation observed in extant Microchiroptera. Because an enlarged cochlea is derived and is associated with use of sophisticated echo- 1998 57SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS location, 3 these observations led Novacek (1985a, 1987, 1991) to conclude that Icaronycteris and Palaeochiropteryx were echolocating bats closely related to extant Microchiroptera. In another study of relative cochlear size in bats, Habersetzer and Storch (1992) suggested that Novacek’s (1985a, 1987, 1991) 3 We recognize two categories of echolocation systems based on the information that is obtained from resulting echoes. We use the term ‘‘sophisticated echolocation’’ to describe systems that can be used for orientation in cluttered environments (e.g., within vegetation or under the forest canopy) and for detection, tracking, and evaluation of moving objects including prey. Calls associated with sophisticated echolocation are produced in the larynx and are often (but not always) highly structured in terms of call length, frequency, and type of modulation. Call structure, duration, and pulse interval may be intentionally varied by the animal depending on the circumstances (e.g., searching for prey, approaching prey, attacking prey), and call structure is often species specific. There is a strong correlation between call type and flight/foraging strategies, in part because the information content of returning echoes varies depending on call structure and pulse interval. However, all sophisticated echolocators apparently record and process information of several types, including relative timing of pulse and echo (including different arrival times at the right and left ears), differences in intensity between pulse and echo, and (in some cases) frequency shifts between pulse and echo. Sophisticated echolocation apparently provides the animal with a detailed acoustic map of its environment, a map that changes moment by moment depending on movement of both the echolocator and the objects in its environment. This form of echolocation occurs in all extant microchiropterans but not in megachiropterans. In contrast, we use the term ‘‘primitive echolocation’’ to refer to less complex systems in which echoes of broadband clicks produced by the animal (by tongueclicks, voice, or some other means) are used to obtain general information about the animals surroundings. Pulses of sound used in primitive echolocation systems are always broadband, low in intensity, short in duration, and never exhibit structured changes in frequency over time. The data used by the animal appears to consist only of differences in timing of pulse and returning echo. The information obtained using primitive echolocation is probably limited to detection of relatively large, stationary obstacles (e.g., cave walls); this form of echolocation is not useful for detecting and tracking prey or orientation in cluttered environments. See discussion under ‘‘Evolution of Echolocation and Foraging Strategies’’ for references and more information about the taxonomic distribution of different echolocation systems. methods were biased by his use of skull length as a measure of body size, because skull length also reflects dietary habits and structure of the dentition. Habersetzer and Storch alleviated this problem by using basicranial width (measured between the outermost bony margins of the semicircular canals) as an indicator of size, and also adopted a more precise method of measuring cochlear width. Using radiographs, they measured cochlear width from the end of the first half turn of the cochlea to the end of the second half turn, thus guaranteeing homologous measurements that excluded the promontorium and vestibular system. Using these methods, Habersetzer and Storch obtained results different from those of Novacek (1985a, 1987, 1991). Rather than having completely nonoverlapping distributions, they found that Microchiroptera (represented by more than 200 species) and Megachiroptera (more than 70 species) exhibited a narrow zone of overlap in relative cochlear size (fig. 29). Some microchiropterans with unusually small cochleae for their body sizes (e.g., Phyllostomus hastatus,Leptonycteris nivalis,Carollia perspicillata,Megaderma lyra,Megaderma spasma) fell within or on the border of the smallest polygon containing nonecholocating megachiropterans (Habersetzer and Storch, 1992). The observed pattern of distribution (fig. 29) was interpreted as evidence that cochlear size in bats varies essentially continuously from Megachiroptera (with the smallest cochleae, lacking sophisticated echolocation) through Rhinolophidae (with the largest cochlea and highly sophisticated echolocation), with cochlear size mirroring the functional significance of the acoustic sense. Habersetzer and Storch (1992) also measured cochlear and basicranial width in Archaeonycteris trigonodon,A.pollex,Palaeochiropteryx tupaiodon,P.spiegeli,Hassianycteris messelensis, and H.revilliodi, and compared the resulting measurements with data from extant forms. They found that both species of Archaeonycteris had a relatively small cochlea, falling in the zone of overlap between Megachiroptera and Microchiroptera, close to the point representing Megaderma spasma. In contrast, both species of Palaeochiropteryx and both species of Hassi- 58 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY Fig. 29. Plot of the oblique diameter (width) of the cochlea versus basicranial width; see text for discussion. This graph was redrawn from Habersetzer and Storch (1992: fig. 4) and several new data points were added. The zone of overlap between Megachiroptera and Microchiroptera is enclosed with a dashed line. Numbered data points refer to the following taxa: 1, Icaronycteris index;2,Archaeonycteris trigonodon;3,Archaeonycteris pollex;4,Palaeochiropteryx tupaiodon;5,Palaeochiropteryx spiegeli;6,Hassianycteris revilliodi;7,Hassianycteris messelensis;8,Rousettus aegypticus;9,Rousettus leschenaulti; 10, Rhinopoma microphyllum; 11, Rhinopoma hardwickei; 12, Craseonycteris thonglongyai; 13, Megaderma lyra; 14, Megaderma spasma; 15, Nycteris grandis; 16, Nycteris macrotis; 17, Nycteris thebaica; 18, Nycteris hispida; 19, Phyllostomus hastatus; 20, Trachops cirrhosus; 21, Leptonycteris nivalis; 22, Carollia perspicillata; 23, Desmodus rotundus; 24, Noctilio albiventris; 25, Mormoops megalophylla; 26, Mystacina robusta; 27, Myzopoda aurita; 28, Thyroptera tricolor; 29, Antrozous pallidus; 30, Tomopeas ravus; 31, Kerivoula pelucida; 32, Tupaia glis; 33, Tupaia javanica. Values for Cynocephalus variegatus (not shown on graph): basicranial width 5 28.1; cochlear width 5 3.85. 1998 65SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS joints are lacking. T1 and T2 are not fused in either outgroup, nor are they fused in most other mammals. In this context, lack of T1– T2 fusion in Icaronycteris and Palaeochiropteryx apparently represents the primitive condition. We were unable to determine the condition of T1 and T2 in Archaeonycteris and Hassianycteris; these taxa are therefore scored ‘‘?’’ for this character. Character 80: Anterior ribs not fused to vertebrae (0); or first rib fused to vertebrae (1); or at least first five ribs fused to vertebrae (2). The anterior ribs are not fused to vertebrae in Icaronycteris,Hassianycteris, and Palaeochiropteryx. Among extant bats, a similar condition is seen in Pteropodidae, Emballonuridae, Rhinopomatidae, Nycteridae, Megadermatidae, and Yangochiroptera. In contrast, the first rib is fused to the vertebral column in Craseonycteridae, and at least the first five ribs are fused to vertebrae in Rhinolophidae. No fusion is seen between ribs and vertebrae in the outgroups and most other mammalian groups. This suggests that the lack of fusion seen in the Eocene bats is primitive. We were unable to determine if rib fusion was present or absent in Archaeonycteris; this taxon is therefore scored ‘‘?’’ for this character. Character 81: Width of first rib similar to other ribs (0); or first rib at least twice the width of other ribs (1). The width of the first rib is similar to that of the other ribs in Icaronycteris and Palaeochiropteryx. This condition is also seen in most extant bats. In contrast, the first rib is at least twice the width of the other ribs in Rhinolophoidea. The width of the first rib is similar to that of the other ribs in both outgroups and most other mammals, suggesting that this condition, which is seen in the Eocene bats, is primitive. We were unable to determine the width of the first rib in Archaeonycteris and Hassianycteris; these taxa are therefore scored ‘‘?’’ for this character. Character 82: First costal cartilage not ossified or fused with manubrium or first rib (0); or first costal cartilage ossified and fused to manubrium (where it appears to form a winglike lateral process of the manubrium) and fused to first rib (1). The manubrium is the anteriormost element of the sternum. It lies at the junction of the proximal clavicles (with which it articulates anterolaterally), the costal cartilages of rib 1 (with which it articulates posterolaterally), and the mesosternum (with which it articulates posteriorly). The manubrium extends laterally only to the level of the clavicular joint in Icaronycteris (fig. 22), Archaeonycteris (fig. 3), Hassianycteris, and Palaeochiropteryx, and there is no evidence of fusion of the first costal cartilage to the manubrium or first rib. This is similar to the condition seen in most extant bats. In contrast, the manubrium has a winglike projection on each side that extends laterally well beyond the clavicular joint in Rhinopomatoidea, Megadermatidae, and Rhinolophidae. In these forms, this winglike lateral projection is fused directly to the first rib. It seems likely that these manubrial ‘‘wings’’ are formed by ossification and fusion of the first costal cartilage (which is typically broad and has a similar shape) with the body of the manubrium. Manubrial wings are lacking and the first rib is not fused to the manubrium or ribs in both outgroups and most mammals, suggesting that the condition seen in the Eocene bats is primitive. Character 83: Second costal cartilage articulates with sternum at manubrium–mesosternum joint (0); or second rib articulates with manubrium, no contact between rib (or costal cartilage) and mesosternum (1). The second costal cartilage articulates with the sternum at the manubrium–mesosternum joint in Icaronycteris (figs. 21, 22), Archaeonycteris (fig. 3) Hassianycteris, and Palaeochiropteryx. This condition is similar to that seen in most extant bats. In contrast, in Rhinolophoidea the second rib articulates with the manubrium and there is no contact between rib (or costal cartilage) and mesosternum. The second costal cartilage articulates with the sternum at the manubrium–mesosternum joint in both outgroups. This suggests that the condition seen in the Eocene bats is relatively primitive. Character 84: Second rib articulates with sternum via costal cartilage (0); or second rib fused to sternum, costal cartilage absent or ossified (1). The second rib articulates with the sternum via a costal cartilage in Icaronycteris (figs. 21, 22), Archaeonycteris (fig. 3), Hassianycteris, and Palaeochirop- 66 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY Fig. 30. Icaronycteris index (holotype; PU 18150). Stereophotographs of close-up dorsal view of the shoulder region. From Jepsen (1970: fig. 9); reprinted from original negatives. teryx. This condition is similar to that seen in most extant bats. In contrast, in Rhinolophidae the second rib is fused to the sternum and the costal cartilage is absent or ossified. In these forms, an extensive sheet of thin bone runs between the second rib and the lateral process of the manubrium. The second rib articulates with the sternum via a costal cartilage in both outgroups and most other mammals, suggesting that the condition seen in the Eocene bats is relatively primitive. Character 85: Mesosternum articulates with at least five costal cartilages posterior to second rib (0); or articulates with four costal cartilages posterior to second rib (1); or articulates with only three costal cartilages posterior to second rib (2). The mesosternum articulates with at least five costal cartilages posterior to the second rib in Icaronycteris (figs. 21, 22), Archaeonycteris (fig. 3), Hassianycteris, and Palaeochiropteryx. Among extant lineages, a similar condition is seen in some Pteropodidae, some Phyllostomidae, Mormoopidae, Noctilionidae, Mystacinidae, Myzopodidae, Natalidae, Molossidae, some Vespertilioninae, and some Myotinae. In contrast, the mesosternum articulates with only four costal cartilages posterior to the second rib in some Pteropodidae, Craseonycteridae, Rhinolophoidea, and some Phyllostomidae. This number is further reduced to three costal cartilages in some Vespertilioninae, Murininae, and Kerivoulinae. The mesosternum articulates with five or more costal cartilages in both outgroups, suggesting that the condition seen in the Eocene bats is relatively primitive. Character 86: Ribs with no anterior laminae (0); or anterior laminae present (1). Anterior laminae are thin plates of bone that run along the leading edges of ribs anterior to the main body of the rib in some bats. These structures, which are often nearly transparent, appear to provide an increased area for muscle attachment. Anterior laminae are absent in Icaronycteris (figs. 22, 30), but are present in Palaeochiropteryx. Among extant forms, anterior laminae are absent in some Emballonuridae, Rhinolophidae, some Phyllostomidae, Mystacinidae, Antrozoidae, 1998 67SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS Fig. 30 (continued). Key to sterophotographs on preceding page. AB, axillary border of scapula; AC, acromion process of scapula; CL, clavicle; HE, head of humerus; HU, humerus; IF, infraspinous fossa of scapula; SF, supraspinous fossa of scapula; SN, suprascapular notch of scapula; T, trochiter. Tomopeatinae, and some Molossinae. Anterior laminae are present in some Pteropodidae, some Emballonuridae, Rhinopomatoidea, Nycteridae, Megadermatidae, some Phyllostomidae, Mormoopidae, Noctilionidae, Nataloidea, some Molossinae, and Vespertilionidae. Both outgroups lack anterior laminae, suggesting that absence of anterior laminae in Icaronycteris is relatively primitive, while presence in Palaeochiropteryx is relatively derived. We were unable to adequately determine presence/absence of these structures in Archaeonycteris and Hassianycteris; these taxa are therefore scored ‘‘?’’ for this character. Character 87: Anterior laminae on ribs narrow, lamina width less than that of main body of rib (0); or anterior laminae wide, equal to or wider than main body of rib (1). The anterior laminae on the ribs of bats may be either relatively narrow (lamina width less than that of the main body of the rib) or wide (equal to or wider than the main body). The anterior laminae are narrow in Palaeochiropteryx. Among extant bats, narrow anterior laminae are found in Pteropodidae, Emballonuridae, Craseonycteridae, Phyllostomidae, Myzopodidae, Molossidae, Vespertilioninae, Myotinae, Murininae, and Kerivoulinae. Wide anterior laminae are seen in Rhinopomatidae, Nycteridae, Megadermatidae, Mormoopidae, Noctilionidae, Thyropteridae, Furipteridae, Natalidae, and Miniopterinae. This character cannot be evaluated for taxa that lack anterior laminae (state ‘‘0’’ of character 86 above); these forms (including Icaronycteris) are scored ‘‘-’’ for this character. Because both outgroups lack anterior laminae, the polarity of this character cannot be determined a priori. As noted above, we were unable to adequately determine presence/absence of anterior laminae in Archaeonycteris and Hassianycteris; these taxa are therefore scored ‘‘?’’ for this character. Character 88: Ribs with no posterior laminae (0); or posterior laminae present (1). Posterior laminae are thin plates of bone that run along the trailing edges of ribs posterior to the main body of the rib. Like the anterior laminae, these structures may be nearly transparent, and they appear to provide an increased area for muscle attachment. Posterior laminae are present in Icaronycteris (figs. 22, 30), Hassianycteris (fig. 4), and Palaeochiropteryx (fig. 2), and they are also present in most extant bats. Posterior laminae are absent in Mystacinidae and some Molossinae. Posterior laminae are also absent in both outgroups, suggesting that their presence in Icaronycteris and Palaeochiropteryx is derived. We were unable to adequately determine presence/absence of these structures in Archaeonycteris; this taxon is therefore scored ‘‘?’’ for this character. Character 89: Posterior laminae on ribs narrow, lamina width less than that of main body of rib (0); or posterior laminae wide, equal to or wider than main body of rib (1). Like the anterior laminae, posterior laminae may be either relatively narrow (lamina width less than that of the main body of the rib) or wide (equal to or wider than the main body). The posterior laminae are narrow in 68 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY Icaronycteris (figs. 22, 30) but are wide in Hassianycteris and Palaeochiropteryx. Among extant bats, narrow posterior laminae are found in Pteropodidae, some Emballonuridae, Craseonycteridae, Phyllostomidae, Noctilionidae, Natalidae, Molossoidea, and some Vespertilioninae. Wide posterior laminae are seen in some Emballonuridae, Rhinopomatidae, Rhinolophoidea, Mormoopidae, Myzopodidae, Thyropteridae, Furipteridae, some Vespertilioninae, Miniopterinae, Myotinae, Murininae, and Kerivoulinae. This character cannot be evaluated for taxa that lack posterior laminae (state ‘‘0’’ of character 88 above); these forms are scored ‘‘-’’ for this character. Because both outgroups lack posterior laminae, polarity of this character cannot be determined a priori. As noted above, we were unable to adequately determine presence/absence of anterior laminae in Archaeonycteris and Hassianycteris; these taxa are scored ‘‘?’’ for this character. Character 90: Anterior face of manubrium small (0); or broad, defined by elevated ridges (1). In addition to its role in connecting the body of the sternum to the clavicles and anterior ribs, the manubrium is also an important site of muscle origin for parts of the m. pectoralis complex, which provides the majority of the power for the downstroke of the wings (Vaughan, 1959, 1970b; Norberg, 1970; Strickler, 1978; Hermanson and Altenbach, 1983, 1985). The anterior face of the manubrium provides the point of origin for part of the anterior division of m. pectoralis, which rotates the humerus and helps to pull it downward and forward and during the downstroke when the wing is protracted against the force of the airstream (Vaughan, 1959, 1970b; Norberg, 1970; Strickler, 1978; Hermanson and Altenbach, 1983, 1985). The anterior face of the manubrium of the sternum is relatively small and poorly defined in Archaeonycteris,Hassianycteris, and Palaeochiropteryx. This condition is similar to that seen in most extant bats. In contrast, the anterior face of the manubrium is a broad, triangular surface that extends onto the lateral processes and is defined by three elevated ridges in some Pteropodidae, Megadermatidae, Rhinolophidae, some Phyllostomidae, Natalidae, and some Molossinae. Among the outgroups, the anterior face of the manubrium is small in Scandentia but is broad and triangular in Dermoptera. Lack of agreement between the outgroups precludes a priori determination of the primitive condition for this character. We were unable to determine the form of the anterior face of the manubrium in Icaronycteris; this taxon is therefore scored ‘‘?’’ for this character. Character 91: Ventral process of manubrium absent (0); or ventral process present, distal tip blunt or rounded (1); or ventral process present, distal tip laterally compressed (2). The ventral process of the manubrium provides the anterior attachment point for a series of ligamentous sheets that run down the midline of the sternum (Vaughan, 1959, 1970b; Norberg, 1970, 1972a; Strickler, 1978; Hermanson and Altenbach, 1983, 1985). The ventral process, together with these ligamentous sheets, forms the origin for the m. pectoralis complex (Vaughan, 1959, 1970b; Norberg, 1970; Strickler, 1978; Hermanson and Altenbach, 1983, 1985). The distal tip of the ventral process of the manubrium is blunt and somewhat rounded in Icaronycteris (figs. 21, 22) and Archaeonycteris (fig. 3). The ventral process in these forms is triangular in cross section and points somewhat posteriorly. In contrast, the tip of the ventral process is laterally compressed (keel-like) in Hassianycteris and Palaeochiropteryx. In cross section, the ventral process in these forms is lens-shaped with the long axis running anteroposteriorly. Among extant bats, a ventral process with a blunt or rounded tip is found in Rhinopomatidae, Megadermatidae, Rhinolophinae, some Hipposiderinae, some Phyllostominae, Mystacinidae, Antrozoidae, Tomopeatinae, some Molossinae, and some Vespertilioninae. A laterally compressed ventral process is seen in Pteropodidae, Emballonuridae, Craseonycteridae, Nycteridae, some Hipposiderinae, some Phyllostominae, Mormoopidae, Noctilionidae, Nataloidea, some Molossinae, some Vespertilioninae, Miniopterinae, Myotinae, Murininae, and Kerivoulinae. Among the outgroups, a blunt ventral process occurs in Dermoptera, but the ventral process is absent in Scandentia. Although the outgroup evidence is ambiguous, this distribution suggests that a ventral process with a blunt, rounded tip (as in Icaronycteris and 1998 69SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS Archaeonycteris) may be primitive, while the laterally compressed, keel-like condition seen in Hassianycteris and Palaeochiropteryx appears to be derived. Character 92: Angle between axis of ventral process and body of manubrium acute (0); or approximately 90 8 (1); or obtuse (2); or ventral process bilobed with one acute and one obtuse process (3). The axis of the ventral process is defined as the long axis of the thickened base and central body of the process. The orientation of the axis of the ventral process (as seen in lateral view) varies among bats. In some taxa, the angle between the ventral process and the body of the manubrium is acute, and the ventral process appears to project posteroventrally. In other forms, the angle is approximately 90 8 (so that the ventral process projects ventrally) or obtuse (ventral processes projects anteroventrally). Although a full range of variation between these conditions is theoretically possible, we found that most species could be easily placed in one of the categories defined above. The only exceptions were a few taxa that have a bilobed ventral process with one acute and one obtuse lobe. Our examinations of the Eocene fossils indicated that the ventral process is obtuse in Icaronycteris (figs. 21, 22) and Palaeochiropteryx, and is oriented at approximately 90 8 in Archaeonycteris (fig. 3). Among extant forms, an obtuse ventral process is seen in some Pteropodidae, some Emballonuridae, Noctilionidae, some Mormoopidae, some Mystacinidae, Myzopodidae, Molossoidea, and Vespertilioninae. The ventral process is oriented at approximately 90 8 in some Pteropodidae, some Emballonuridae, Rhinopomatoidea, some Phyllostomidae, some Mormoopidae, Thyropteridae, Myotinae, Murininae, and Kerivoulinae. An acute ventral process occurs in Megadermatidae, Rhinolophidae, some Phyllostomidae, and Natalidae. The ventral process is bilobed in Nycteridae, Furipteridae, and Miniopterinae. Among the outgroups, an obtuse ventral process occurs in Dermoptera. This feature cannot be evaluated in taxa that lack a ventral process on the manubrium (state ‘‘0’’ of character 91 above). Scandentia, whose members lack a ventral process, is therefore scored ‘‘-’’ for this character. Although the outgroup evidence is ambiguous, presence of an obtuse ventral process in Dermoptera suggests that this condition, which is also seen in Icaronycteris and Palaeochiropteryx, may be primitive. The approximate 90 8 angle of the ventral process in Archaeonycteris would thus be interpreted as a derived condition. We were unable to determine the angle of the ventral process in Hassianycteris; this taxon is therefore scored ‘‘?’’ for this character. Character 93: Length of manubrium posterior to lateral processes . 2.5 times transverse width (0); or length , 2 times transverse width (1). The length of the manubrium posterior to the lateral processes is less than twice the transverse width of this portion of the manubrium in Icaronycteris (figs. 21, 22), Archaeonycteris (fig. 3), Hassianycteris, and Palaeochiropteryx. This condition is similar to that seen in most extant bats. In contrast, the manubrium is relatively elongated (posterior portion . 2.5 times transverse width) in Nycteridae, some Phyllostomidae, and Molossidae. Both conditions occur among the outgroups—the manubrium is relatively short in Dermoptera but is elongate in Scandentia. Accordingly, the primitive condition of this feature cannot be determined a priori. Character 94: Mesosternum narrow, mean width less than half the distance between clavicles at sternoclavicular joint (0); or mesosternum broad, mean width greater than three-fourths the distance between clavicles (1). The mesosternum ( 5 body of sternum) articulates with the manubrium anteriorly and with the xiphisternum posteriorly. The relative width of the mesosternum in a bat can be estimated by comparing mean mesosternal width with the transverse distance between the right and left clavicles at their joints with the manubrium. 4 The mesosternum is relatively narrow (mesosternal width less than half the interclavicular distance) in Icaronycteris (figs. 21, 22), Archaeonycteris, 4 Modifications of the manubrium and ribs in a number of taxa (e.g., rhinolophoids) preclude meaningful size comparisons of the mesosternum with these elements. Accordingly, we chose to compare width of the mesosternum with the interclavicular distance because our observations suggest that the distance between the clavicles is correlated principally with body size. 70 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY Hassianycteris, and Palaeochiropteryx. This is similar to the condition seen in Pteropodidae, Emballonuridae, Rhinopomatoidea, Megadermatidae, Rhinolophidae, Noctilionoidea, Mystacinidae, Myzopodidae, Molossoidea, and some Vespertilionidae. In contrast, the mesosternum is broad (width greater than three-fourths the interclavicular distance) in Nycteridae, Thyropteridae, Furipteridae, Natalidae, some Vespertilioninae, Miniopterinae, Myotinae, Murininae, and Kerivoulinae. The mesosternum is relatively narrow in both outgroups, suggesting that the narrow condition seen in the Eocene bats is relatively primitive. Character 95: Xiphisternum without keel (0); or with prominent median keel (1). The xiphisternum lacks a median longitudinal keel on its ventral surface in Icaronycteris (fig. 22), Archaeonycteris, and Palaeochiropteryx. This condition is seen among extant bats in Pteropodidae, Rhinopomatidae, Hipposiderinae, some Phyllostomidae, Noctilionidae, Myzopodidae, Molossoidea, and some Vespertilioninae. In contrast, a prominent, ventrally projecting median keel is present on the xiphisternum in Emballonuridae, Craseonycteridae, Nycteridae, Megadermatidae, Rhinolophidae, some Phyllostomidae, Mormoopidae, Mystacinidae, Thyropteridae, Furipteridae, Natalidae, some Vespertilioninae, Miniopterinae, Myotinae, Murininae, and Kerivoulinae. A xiphisternal keel is absent in both outgroups, suggesting that absence of a keel is relatively primitive. We were unable to determine if a xiphisternal keel is present in Hassianycteris due to distortion and flattening of the sternal region in all available specimens; this taxon is therefore scored ‘‘?’’ for this character. Character 96: Posterior xiphisternum with wide lateral flare (0); or not laterally flared (1). The posterior xiphisternum has a wide lateral flare in Icaronycteris (fig. 22), Hassianycteris, and Palaeochiropteryx. This flare is produced by a steady increase in width of the xiphisternum from the anterior to the posterior end of the element. In contrast, the xiphisternum is not flared (and width of the posterior end is approximately equal to the width of the anterior end) in Archaeonycteris. Among extant forms, a lateral flare is present in some Pteropodidae, some Emballonuridae, Rhinopomatoidea, Nycteridae, Megadermatidae, Hipposiderinae, Noctilionoidea, Mystacinidae, Thyropteridae, Molossoidea, and Vespertilionidae. There is no lateral flare in some Pteropodidae, some Emballonuridae, Rhinolophinae, Myzopodidae, and Furipteridae. Both outgroups have a xiphisternum with a wide lateral flare. This pattern suggests that the flared morphology seen in Icaronycteris,Hassianycteris, and Palaeochiropteryx may be relatively primitive, while absence of a xiphisternal flare in Archaeonycteris may represent a relatively derived condition. P ECTORAL G IRDLE Character 97: Acromion process without medial shelf (0); or with shelf that projects medially over supraspinous fossa or medial base of acromion process (1). The distal half of the medial surface of the acromion process serves as the attachment point for one end of the transverse scapular ligament ( 5 dorsal scapular ligament), which stretches between the acromion and the anteromedial rim of the scapula dorsal to m. supraspinatus (Vaughan, 1959, 1970b; Norberg, 1970; Strickler, 1978; Hermanson and Altenbach, 1983, 1985). In various bats the transverse scapular ligament provides an increased attachment area for m. supraspinatus (which elevates, extends, and rotates the humerus), m. acromiodeltoideus (which elevates the humerus), and m. spinodeltoideus (which elevates and flexes the humerus; Vaughan, 1959, 1970b; Norberg, 1970; Strickler, 1978; Hermanson and Altenbach, 1983, 1985). Presence of a medial shelf on the distal acromion is probably correlated with changes in the extent and orientation of the transverse scapular ligament and in the origins of those parts of m. supraspinatus and m. acromiodeltoideus that originate directly from the acromion process. The acromion process of the scapula lacks a medial shelf in Icaronycteris,Archaeonycteris,Hassianycteris, and Palaeochiropteryx. This is similar to the condition seen in most extant bats. In contrast, the acromion process has a shelflike projection that extends medially over the supraspinous fossa or medial base of the acromion process in Thyropteridae, Furipteridae, Natalidae, some Vesperti- 1998 71SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS lioninae, Miniopterinae, and Kerivoulinae. No medial shelf is seen in either outgroup, suggesting that absence of a medial shelf on the acromion process is the primitive condition. Character 98: Tip of acromion process without anterior projection (0); or with triangular anterior projection (1). The tip of the acromion process lacks an anterior projection in Icaronycteris,Hassianycteris, and Palaeochiropteryx. The acromion similarly lacks such a projection in Pteropodidae, Emballonuridae, Yinochiroptera, Phyllostomidae, Natalidae, and Murininae. In contrast, a triangular anterior projection is present just ventral and medial to the tip of the acromion process in Mormoopidae, Noctilionidae, Mystacinidae, Myzopodidae, Thyropteridae, Furipteridae, Molossoidea, Vespertilioninae, Miniopterinae, Myotinae, and Kerivoulinae. This projection curves slightly ventrally in some taxa. No anterior projection is present on the acromion process in the outgroups; this suggests that absence of an anteroventral projection in the Eocene forms may be primitive. We were not able to determine if an anteroventral process is present in Archaeonycteris; this taxon is therefore scored ‘‘?’’ for this character. Character 99: Distal acromion process without posterolateral projection (0); or with triangular posterolateral projection (1). The entire lateral surface of the acromion process serves as part of the origin of m. acromiodeltoideus, a muscle that elevates and rotates the humerus and provides important control during the upstroke of the wing (Vaughan, 1959, 1970b; Strickler, 1978). Presence of a posterolateral projection on the distal acromion may serve to increase the area of attachment for m. acromiodeltoideus, and may affect the moment arm for at least some of the fibers in this muscle. The acromion process lacks a posterolateral projection in Icaronycteris,Archaeonycteris,Palaeochiropteryx, and some specimens of Hassianycteris (e.g., SMF Me 1500). This condition is similar to that seen in most extant bats. In contrast, a triangular posterolateral projection is present on the distal acromion process in extant Molossidae and some specimens of Hassianycteris (e.g., SMF Me 1540a). Both specimens of Hassianycteris are similar in size; accordingly, this variation may represent either within-species polymorphism (within H.messelensis)or taxonomic polymorphism (it may differentiate H.messelensis from a previously unidentified specimen of H.revilliodi). We are unaware of any within-species polymorphism in this character in extant bat species, so we favor the latter interpretation. No posterolateral projection is present on the acromion process in the outgroups, suggesting that absence of a posterolateral projection in the Eocene bats is primitive. Presence of a triangular posterolateral projection on the acromion in some Hassianycteris is apparently a derived condition. Character 100: Dorsal articular facet (for trochiter of humerus) absent from scapula (0); or present (1). A secondary articulation between the humerus and scapula occurs in many bats when the humerus is abducted and an enlarged trochiter ( 5 greater tuberosity) contacts a dorsal articular facet on the scapula. As discussed by Vaughan (1959, 1970b), Hill (1974), Strickler (1978), Hill and Smith (1984), Altenbach and Hermanson (1987), and Schlosser-Strum and Schleimann (1995), this secondary articulation forms a critical part of a ‘‘locking mechanism’’ in the shoulder. Functional explanations for the shoulderlocking mechanism are diverse and, in some cases, contradictory (see discussion in Schlosser-Strum and Schleimann, 1995). One possibility is that this mechanism functions in part to arrest the upstroke of the wing, thus facilitating long-range, fast flight (1) by providing a longer recovery period for critical flight muscles between wing beats, and (2) by reducing the amount of force that must be exerted by many of the same muscles (Vaughan, 1959, 1970b; Hill, 1974; Strickler, 1978; Hill and Smith, 1984). An alternative view based on new morphological and experimental evidence suggests that the secondary shoulder joint serves to increase the moment arm of m. pectoralis and reduce pronatory movements of the abducted forearm during the downstroke (Altenbach and Hermanson, 1987; Schlosser-Strum and Schliemann, 1995). The dorsal articular facet is situated on the dorsal surface of the scapula immediately anteromedial to the glenoid fossa. Icaronycteris 72 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY lacks a dorsal articular facet. In contrast, a dorsal articular facet is present in Hassianycteris and Palaeochiropteryx. Most extant bats have a dorsal articular facet; exceptions include Pteropodidae, some Emballonuridae, Rhinopomatidae, and Noctilionidae. Both outgroups and most other mammals do not have a secondary articulation between the humerus and scapula, and therefore lack a dorsal articular facet on the scapula. This suggests that absence of a dorsal articular facet in Icaronycteris represents a relatively primitive condition. We were unable to determine if a dorsal articular facet occurs in Archaeonycteris; this taxon is therefore scored ‘‘?’’ for this character. Character 101: Dorsal articular facet faces dorsolaterally and consists of small groove on anteromedial rim of glenoid fossa (0); or faces dorsolaterally and consists of an oval facet on anteromedial rim of glenoid fossa (1); or faces dorsally and consists of a large, flat facet clearly separated from glenoid fossa (2).InPalaeochiropteryx, the dorsal articular facet consists of a small, dorsolaterally facing groove on the anteromedial rim of the glenoid fossa of the scapula. In contrast, the dorsal articular facet consists of an oval, dorsolaterally facing facet on the anteromedial rim of the glenoid fossa in Hassianycteris. Among extant bats, the ‘‘small groove’’ condition is seen in Emballonuridae, Nycteridae, and some Hipposiderinae. An oval, dorsolaterally facing dorsal articular facet on the anteromedial rim of the glenoid fossa is found in Craseonycteridae, some Megadermatidae, some Phyllostomidae, Mormoopidae, Thyropteridae, Furipteridae, and Natalidae. In contrast, the dorsal articular facet is a large, flat facet that faces dorsally and is clearly separated from the glenoid fossa in some Megadermatidae, Rhinolophinae, some Hipposiderinae, some Phyllostomidae, Mystacinidae, Myzopodidae, Molossoidea, and Vespertilionidae. This character cannot be scored in taxa that lack a dorsal articular facet or those in which presence/absence of a dorsal articular facet has not been determined. Both outgroups, Pteropodidae, and Icaronycteris lack a dorsal articular facet; these taxa are scored ‘‘-’’ for this character. Archaeonycteris is scored ‘‘?’’ to reflect insufficient information for this taxon. This character cannot be polarized a priori because both outgroups lack a dorsal articular facet. Character 102: Infraspinous fossa of scapula narrow, length $ 2 times width (0); or wide, length # 1.5 times width (1). The infraspinous fossa provides the site of origin for m. infraspinatus and m. teres major, muscles that act to flex, rotate, and (in the case of m. infraspinatus) abduct the humerus (Vaughan, 1959, 1970b; Norberg, 1970; Strickler, 1978; Hermanson and Altenbach, 1983, 1985). The width of the infraspinous fossa is measured from the junction of the scapular spine and the vertebral border of the scapula to the axillary border in a line perpendicular to the long axis of the fossa (which is defined by the line of maximum length). The infraspinous fossa is narrow (length greater than or equal to twice the width) in Icaronycteris (figs. 22, 30), Archaeonycteris, and Hassianycteris. In contrast, the infraspinous fossa is relatively wide (length # 1.5 times width) in Palaeochiropteryx (fig. 2). Among extant forms, a narrow infraspinous fossa is seen in Pteropodidae, Rhinopomatoidea, Rhinolophoidea, some Mormoopidae, Myzopodidae, Thyropteridae, some Molossinae, and Murininae. A wide infraspinous fossa occurs in Phyllostomidae, Noctilionidae, some Mormoopidae, Mystacinidae, Furipteridae, Natalidae, Antrozoidae, Tomopeatinae, some Molossinae, Vespertilioninae, Miniopterinae, Myotinae, and Kerivoulinae. The infraspinous fossa in both outgroups is narrow, suggesting that a narrow fossa is relatively primitive and that the wide infraspinous fossa in Palaeochiropteryx represents a derived condition. Character 103: Infraspinous fossa with one facet (0); or two facets (2); or three facets (2). Faceting of the infraspinous fossa, which effectively serves to increase the surface area without increasing the outline dimensions of the fossa, is thought to function to increase the area of origin for m. infraspinatus and m. subscapularis (Vaughan, 1959; see discussion of m. subscapularis function below under character 106). Faceting may also reflect compartmentalization or subdivision of fibers of these muscles into units with distinct functions, although this has yet to be investigated. 1998 73SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS The infraspinous fossa has two facets in Icaronycteris,Archaeonycteris,Hassianycteris, and Palaeochiropteryx. The smaller of these (the medial facet) lies posterior to the scapular spine and faces posterolaterally; the larger lateral facet, which is convex, lies more posteriolaterally and faces dorsally. Among extant bats, two facets are found in some Pteropodidae, Rhinopomatoidea, Nycteridae, some Phyllostomidae, Mystacinidae, Furipteridae, Natalidae, Antrozoidae, Tomopeatinae, some Vespertilioninae, and Myotinae. In contrast, the infraspinous fossa has three facets in some Pteropodidae, Emballonuridae, Megadermatidae, Rhinolophidae, some Phyllostomidae, Mormoopidae, Noctilionidae, Myzopodidae, Thyropteridae, Molossinae, some Vespertilioninae, Myotinae, Murininae, and Kerivoulinae. In these forms, the facet just posterior to the scapular spine is called the posteromedial facet; this appears to be homologous to the medial facet found in the same position in bats that have only two facets. The larger lateral facet is further subdivided into an intermediate facet and a posterolateral facet. The posterolateral facet lies along the lateral edge of the scapula and faces posterolaterally; the intermediate facet lies between the postermedial and posterolateral facets and faces anteromedially. The infraspinous fossa is not subdivided in the outgroups. In these forms, there is only a single, large convex facet in the infraspionous fossa. Accordingly, it is not possible to determine a priori the primitive condition for bats (two facets or three facets). Character 105: Lateral or posterolateral facet of infraspinous fossa restricted, does not extend into infraglenoid region anteriorly or wrap around intermediate facet at posterior (caudal) angle of scapula (0); or posterolateral facet more extensive, extends into infraglenoid region and wraps around caudal end of intermediate facet (1).InIcaronycteris,Archaeonycteris,Hassianycteris, and Palaeochiropteryx, the extent of the lateral facet of the infraspinous fossa is relatively restricted; this facet does not extend into the infraglenoid region or wrap around the medial facet at the posterior (caudal) angle of the scapula. A similar condition is seen in most extant bats regardless of the number of facets in the infraspinous fossa. In contrast, the posterolateral facet is more extensive in Megadermatidae and Rhinolophidae. In these forms (which have three facets), the posterolateral facet extends into the infraglenoid region anterolaterally, and it wraps around the caudal end of the intermediate fossa posteriorly. This character cannot be evaluated in taxa that have only one infraspinous facet, which precludes useful scoring of this character in the outgroups; these taxa are scored ‘‘-’’ for this feature. Accordingly, the primitive condition for bats cannot be reconstructed a priori. Character 106: Thick lip present along axillary border of scapula (0); or thick lip with bladelike lateral edge present (1); or thick lip absent, axillary border flat or slightly upturned (2). The axillary border of the scapula provides part of the site of origin of m. subscapularis, which adducts and extends the humerus (although a more important function may be to stabilize and provide fine control of the wing), and m. teres minor, which is a weak flexor of the humerus (Vaughan, 1959, 1970b; Norberg, 1970; Strickler, 1978; Hermanson and Altenbach, 1983, 1985). Morphology of the edge of the scapula along its axillary (lateral/posterolateral) border varies among bats. A thick lip is present along the axillary border of the scapula in Icaronycteris (fig. 30), Archaeonycteris,Hassianycteris, and Palaeochiropteryx. The bone itself is fairly thin; the appearance of a thickened lip is produced by an abrupt fold near the axillary border. This condition is similar to that seen among extant Pteropodidae, some Emballonuridae, Mystacinidae, Thyropteridae, Molossoidea, Vespertilioninae, Miniopterinae, Myotinae, and Kerivoulinae. In contrast, the axillary border of the scapula is characterized by a thick lip with a bladelike lateral edge in some Emballonuridae, Noctilionoidea, and Myzopodidae. No thickening is seen along the axillary border in the remaining lineages of bats; instead, the edge of the scapula is flat or slightly upturned, and the abrupt fold is absent. Both outgroups exhibit a simple, thick lip (state ‘‘0’’) on the axillary border, suggesting that this is the primitive condition. Character 107: Pit for attachment of clavicular ligament absent from scapula (0); or present anterior and medial to glenoid fossa 74 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY (1). The clavicular ligament extends between the base of the coracoid process and the clavicle. There is no evidence of a pit for attachment of the clavicular ligament on the scapula in Icaronycteris and Hassianycteris, and the same is true in most extant bat lineages. In contrast, a distinct pit for the clavicular ligament is present anterior and medial to the glenoid fossa in some Emballonuridae, Rhinolophoidea, some Phyllostomidae, and Mormoopidae. When a dorsal articular facet is present, the pit for the clavicular ligament is located medial to this structure. The pit for the clavicular ligament is absent in both outgroups, suggesting that absence of this structure represents the primitive condition. We were unable to determine if a pit for the clavicular ligament is present in Archaeonycteris and Palaeochiropteryx; these taxa are therefore scored ‘‘?’’ for this character. Character 108: Anteromedial edge of scapula without projections or flanges (0); or with triangular anteromedial flange (1). The anteromedial flange is a roughly triangular flange that projects ventrally from the anteromedial edge of the scapula (medial to the suprascapular notch) in some bats (Vaughan, 1959, 1970a; Norberg, 1970; Strickler, 1978; Hermanson and Altenbach, 1983, 1985). The dorsal aspect of the scapular edge in this region serves as the origin of the transverse scapular ligament, while part of the origins of m. subscapularis and the anterior division of m. serratus anterior occupy its ventral surface (Vaughan, 1959, 1970b; Norberg, 1970; Strickler, 1978; Hermanson and Altenbach, 1983, 1985). Projections from the border of the scapula in this region apparently simultaneously serve to provide an increased area of attachment for the transverse scapular ligament and an increased area for muscle origins. Presence of a large, triangular anteromedial flange is correlated with relatively large size of m. subscapularis (Vaughan, 1959), which is one of the principal adductors and extensors of the humerus. M. subscapularis plays an important role in the downstroke during flight and also helps to support the weight of the anterior part of the body in terrestrial locomotion (Vaughan, 1959, 1970b; Strickler, 1978). The anteromedial rim of the scapula lacks projections or flanges in Icaronycteris,Archaeonycteris, and Hassianycteris. In contrast, a triangular anteromedial flange is present on the medial superior (anterior) border of the scapula in Palaeochiropteryx. Among extant bats, the anteromedial flange is absent in Pteropodidae, Emballonuridae, Rhinopomatoidea, Yinochiroptera, Noctilionoidea, Mystacinidae, Myzopodidae, Thyropteridae, and Furipteridae. A large, triangular anteromedial flange is present in Myzopodidae, Natalidae, Molossoidea, and Vespertilionoidea. Absence of this flange in both outgroups (and most other mammals) indicates that absence of this structure is the primitive condition. Character 109: Coracoid process stout and of moderate length (0); or very long and thin (1). The coracoid process of the scapula is the site of origin of the coracoid head ( 5 short head) of m. biceps brachii. M. biceps brachii is a two-part muscle that flexes and rotates the forearm, adducts the wing, and helps to hold the forearm rigidly outstretched against the the opposing action of m. triceps during much of the downstroke (Vaughan, 1959, 1970b; Norberg, 1970; Strickler, 1978; Hermanson and Altenbach, 1983, 1985). Length of the coracoid process plays a role in function of m. biceps brachii because a long coracoid process enables the coracoid head of the biceps to act as an adductor of the wing by placing its origin below the line of the long axis of the humerus; the greater the displacement from this axis, the greater the mechanical advantage of the biceps as an adductor (Vaughan, 1959, 1966). The coracoid process of the scapula is stout (wider than the clavicle) and of moderate length in Icaronycteris,Archaeonycteris,Hassianycteris, and Palaeochiropteryx. This condition is similar to that seen in most extant bats. In contrast, the coracoid is very long and thin (distal half not as wide as clavicle) in Furipteridae, Natalidae, Tomopeatinae, and some Molossinae. Both outgroups exhibit the ‘‘short and stout’’ morphology, suggesting that this is the primitive condition for the coracoid process. Character 110: Coracoid process curves ventrolaterally (0); or curves ventrally (1); or curves ventromedially (2). The direction of curvature of the coracoid process varies considerably among bats (fig. 31). Orienta- 1998 81SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS it II is ossified in both outgroups, suggesting that the condition seen in the Eocene bats is primitive. Character 150: Wing digit II with ossified second phalanx (0); or second phalanx unossified or absent (1). The second phalanx is defined by its articulation with the distal end of the first phalanx of a particular digit. The second phalanx in wing digit II is ossified in Icaronycteris,Archaeonycteris,Hassianycteris, and Palaeochiropteryx. Among extant bats, the second phalanx is present and ossified only in Pteropodidae and Rhinopomatidae; this element is unossified or absent in all other lineages. Digit II of the manus has an ossified second phalanx in both outgroups, suggesting that the condition seen in the Eocene bats is primitive. Character 151: Wing digit II with ossified third phalanx (0); or third phalanx unossified or absent (1). The third phalanx ( 5 ungual phalanx) is defined by its articulation with the distal end of the second phalanx, and in mammals it typically bears a claw or homologous structure (nail or hoof). The third phalanx of wing digit II is present and ossified in Icaronycteris and Archaeonycteris.In both taxa, the third phalanx of this digit clearly bore a claw. Indeed, the specific epithet ‘‘index’’ for the type species of Icaronycteris refers to presence of a claw on the index finger (digit II; Jepsen, 1966). In contrast, the third phalanx of digit II is unossified or absent in Hassianycteris and Palaeochiropteryx. Among extant bat lineages, the third phalanx is present and ossified only in Pteropodidae; it is unossified or absent in all other groups. Digit II of the manus has an ossified third phalanx in both outgroups, suggesting that the condition seen in Icaronycteris and Archaeonycteris is primitive. The state seen in Hassianycteris and Palaeochiropteryx is apparently derived. Character 152: Wing digit III with third phalanx completely ossified (0); or third phalanx ossified only at the base (1); or third phalanx unossified or absent (2). The third phalanx of wing digit III is small but fully ossified in Icaronycteris. In contrast, the third phalanx of this digit is either unossified or absent in Archaeonycteris,Hassianycteris, and Palaeochiropteryx. A small but fully ossified third phalanx is found among extant Phyllostomidae, Mormoopidae, Mystacinidae, Myzopodidae, and Thyropteridae. The third phalanx is ossified only at the base (and the distal portion of the element remains cartilaginous) in Furipteridae, Natalidae, Molossoidea, and some Vespertilionidae. The third phalanx is completely unossified or absent in Pteropodidae, Emballonuridae, Yinochiroptera, Noctilionidae, and some Vespertilionidae. The third phalanx in digit III of the manus is fully ossified in both outgroups and most other mammals, suggesting that the condition seen in Icaronycteris is primitive. Lack of an ossified third phalanx in Archaeonycteris,Hassianycteris, and Palaeochiropteryx is relatively derived. P OSTERIOR A XIAL S KELETON AND P ELVIS Character 158: No vertebral fusion in posterior thoracic and lumbar series (0); or at least three vertebrae fused (1). The vertebral column of most mammals contains posterior thoracic and lumbar vertebrae as separate elements that can move relative to one another. They articulate at various hypophyseal joints, but are not fused. This is also true of the posterior thoracic and lumbar series in Icaronycteris (frontispiece, fig. 21), Archaeonycteris,Hassianycteris (fig. 4), and Palaeochiropteryx (fig. 3), all of which exhibit no evidence of vertebral fusion. Among extant bats, vertebral fusion in this area is similarly absent in Pteropodidae, Emballonuridae, Rhinopomatidae, Nycteridae, some Megadermatidae, Rhinolophidae, some Hipposiderinae, Phyllostomidae, Noctilionidae, Mystacinidae, Myzopodidae, Thyropteridae, Molossoidea, and Vespertilionidae. In contrast, at least three vertebrae are fused together in the posterior thoracic and lumbar region in Craseonycteridae, some Megadermatidae, some Hipposiderinae, Mormoopidae, Furipteridae, and Natalidae. Both outgroups lack vertebral fusion in the posterior thoracic and lumbar regions, suggesting that the condition seen in the Eocene bats is relatively primitive. Character 159: Sacrum terminates anterior to acetabulum (0); or extends posteriorly to at least the midpoint of the acetabulum (1). The sacrum (as defined in this study) includes all vertebrae that articulate with the 82 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY pelvis or are fused with those that do. The posterior end of the sacrum terminates anterior to the acetabulum in Hassianycteris.In contrast, the sacrum extends posteriorly to at least the midpoint of the acetabulum in Icaronycteris (fig. 23) and Palaeochiropteryx. 7 Among extant bats, the sacrum terminates anterior to the acetabulum in Nycteridae and Rhinolophidae; it extends posteriorly to a least the midpoint of the acetabulum in all other lineages. The sacrum terminates anterior to the acetabulum in both outgroups, suggesting that the condition seen in Hassianycteris is primitive. The condition seen in Icaronycteris and Palaeochiropteryx appears derived. We were unable to determine the extent of the sacrum in Archaeonycteris; this taxon is scored ‘‘?’’ for this character. Character 160: Sacral laminae narrow or absent, vertebra width (including laminae) less than or equal to three-fourths vertebral body length (0); or laminae broad, vertebra width equal to or greater than vertebral length. Sacral laminae are thin plates of bone that extend laterally from the sacral vertebrae posterior to the iliosacral joint. These structures, which are homologous with the transverse processes of lumbar vertebrae, do not articulate with the pelvis. Instead, they typically form joints with one another that may or may not be fully fused. Gaps are frequently present between successive laminae, and lateral enclosure of such gaps may produce a series of sacral foramina between successive laminae. The sacral laminae are broad in Icaronycteris (fig. 23) and Palaeochiropteryx. Broad sacral laminae are also present in Nycteridae, Myzopodidae, Thyropteridae, Molossoidea, and Vespertilionidae. In these forms, the width of each postischial sacral vertebra (including the sacral laminae) is subequal to or greater than the length of the 7 Our observations do not agree with those of Russell and Sige´ (1970), who reported that the sacrum in Palaeochiropteryx was very small and included only a single vertebra. We attribute this discrepency to an incorrect reconstruction of this region by Russell and Sige´ (1970), who were working with relatively poorly preserved material. Discovery of better preserved specimens since their publication (e.g., HLMD Me 15025) demonstrate that Palaeochiropteryx had a long sacrum composed of multiple vertebrae. vertebral body. In constrast, the sacral laminae are narrow or absent in all other taxa, extending only a short distance laterally from the bodies of the vertebrae. Accordingly, vertebral width is less than or equal to threefourths the length of the body of the vertebra. Among extant bats, the sacral laminae are narrow or are absent in Pteropodidae, Emballonuridae, Rhinopomatoidea, Megadermatidae, Rhinolophidae, Noctilionoidea, Mystacinidae, Furipteridae, and Natalidae. Among the outgroups, the sacral laminae are broad in Scandentia but are absent in Dermoptera. Accordingly, this character cannot be polarized a priori. Morphology of the sacral laminae in unknown in Archaeonycteris and Hassianycteris; these forms are therefore scored ‘‘?’’ for this character. Character 161: Dorsomedial edge of ascending process of ilium upturned, flares dorsally above the level of iliosacral articulation, iliac fossa large and well defined (0); or dorsomedial edge not upturned, does not extend dorsally beyond the level of the iliosacral articulation, iliac fossa not large or well defined (1). The iliosacral articulation in mammals is formed between the ascending process of the ilium and the sacral vertebrae. The dorsomedial edge of the ascending process forms the origin for m. tensor fascia latae, which flexes and abducts the femur (Vaughan, 1959, 1970b). M. gluteus medius (which flexes, abducts, and rotates the femur) also originates from the dorsomedial edge, although the iliac fossa on the dorsolateral surface of the ascending process forms the principal origin of this muscle (Vaughan, 1959, 1970b). The dorsomedial edge of the ascending process of the ilium is upturned and flares dorsally above the level of the iliosacral articulation in Icaronycteris,Hassianycteris, and Palaeochiropteryx. Among extant bats, a dorsomedially flared ascending process is seen only in Rhinolophidae. In all taxa this flare is associated with a large, well-defined iliac fossa. In contrast, the dorsomedial edge of the ascending process is not upturned and the flare is absent in all other bats. The dorsomedial edge terminates at the level of the iliosacral articulation, and the iliac fossa is relatively small and poorly defined. The upturned edge and well-defined iliac fossa are 1998 83SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS present in Scandentia but absent in Dermoptera; accordingly, this character cannot be polarized a priori. We were unable to determine the state of the ilium in Archaeonycteris; this taxon is therefore scored ‘‘?’’ for this character. Character 162: Ischium with large ischial tuberosity that projects dorsally from posterior horizontal ramus (0); or ischial tuberosity small or absent, does not project dorsally beyond level of ramus (1). The dorsal ischial tuberosity is a projection that extends dorsally and/or posteromedially from the posterior end of the ischium, originating from the ‘‘corner’’ of the pelvis at the junction between the horizontal and vertical rami of the ischium (Vaughan, 1959, 1970a). Two muscles may originate from the ischial tuberosity: m. semitendinosus and m. semimembranosus, both of which act to extend the femur and flex the lower leg (Vaughan, 1959, 1970b). Vaughan (1959) recognized an ischial tuberosity in all of the bats that he dissected, but noted that in some forms (e.g., Eumops) the ischial tuberosity is large, projects dorsally above the level of the horizontal ramus, and serves as the site of origin of both muscles. In other taxa (e.g., Myotis,Macrotus), the tuberosity is smaller, projects somewhat medially (not dorsally), and serves as the site of origin for just m. semitendinosus. In the latter taxa, m. semimembranosus originates from the lateral surface of the caudal border of the ischium (Vaughan, 1959). Our examination of skeletons from numerous families revealed a dichotomy between forms with a large, dorsally projecting ischial tuberosity and those in which the ischial tuberosity is small or absent. Although we have no new muscle data, we infer that morphology of the pelvis is correlated with differences in muscle origins as observed by Vaughan (1959). There is no evidence of a large, dorsally projecting ischial tuberosity on the pelvis of Icaronycteris (fig. 23), Hassianycteris, and Palaeochiropteryx. In contrast, a large, dorsally projecting ischial tuberosity is present in Archaeonycteris. Among extant bats, a large dorsal ischial tuberosity is present only in Mystacinidae and Molossidae. A large dorsally projecting ischial tuberosity is present in one outgroup (Scandentia) but not in the other (Dermoptera); accordingly, this character can not be polarized a priori. Character 163: Pubic spine absent (0); or straight (1); or tip of pubic spine bent sharply dorsally (2). The pubic spine is an elongate projection from the anteroventral corner of the pubis (Vaughan, 1959; Walton and Walton, 1970; Simmons, 1994). Presence of a pubic spine is a synapomorphy of Chiroptera (Simmons, 1994). Two muscles originate from the pubic spine in bats: m. pectineus from the base and lateral surface of the spine, m. gracilis from the entire ventrolateral surface of the spine to the tip (Vaughan, 1959, 1970). M. pectineus is an adductor and flexor of the femur, while m. gracilis is a flexor of the lower leg and an adductor of the hind limb (Vaughan, 1959, 1970b). Additionally, the pubic spine is the site of insertion of m. psoas minor, which originates from the lumbar vertebrae (Vaughan, 1959, 1970b). Contraction of m. psoas minor pulls the ventral part of the pelvis forward, thereby arching the posterior lumbar section of the vertebral column. Vaughan (1959) noted that this action may be useful in doubling up the body when the bat is grooming itself while hanging, or when pinning an insect to the uropatagium while adjusting the grip of the jaws on the prey. Action of this muscle may also help to brace the vertebral column when the bat lands, and also while the bat is flying (i.e., against the shock of the airstream hitting the uropatagium during sudden maneuvers; Vaughan, 1959). The pubic spine is relatively straight and points anterodorsally in Icaronycteris (fig. 23), Archaeonycteris, and Palaeochiropteryx. Among extant bats, this condition is seen in Pteropodidae, Emballonuridae, Rhinopomatoidea, Nycteridae, and Yangochiroptera. In these forms, the course of the pubic spine is essentially parallel to that of the ilium, with the long axis of the spine directed at the anterior lumbar vertebrae. In contrast, the tip of the pubic spine is sharply upturned in Megadermatidae, Rhinolophinae, and Hipposiderinae. The tip of the pubic spine in these taxa points dorsally toward the cranial end of the ilium. In Hipposiderinae, a bony extension from the pubic spine contacts the anterior ilium, thereby enclosing a preaceta- 84 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY bular foramen. The pubic spine is absent in both outgroups, so the form of the pubic spine (straight or with tip upturned) cannot be polarized a priori. We could not determine the form of the pubic spine in Hassianycteris, so this taxon is scored ‘‘?’’ for this character. Character 165: Obturator foramen normal, rim well defined (0); or foramen partially infilled with thin, bony sheet along posteroventral rim (1). The obturator foramen of the pelvis is enclosed dorsally and posteriorly by the rami of the ischium and anteriorly and ventrally by the pubis. In most mammals, the rim of the obturator foramen is well defined and rounded in cross section. Muscles originating around the rim of the obturator foramen include m. adductor brevis (extensor and/or adductor of femur), m. adductor magnus (extensor and rotator of femur), and m. obturator externus (extensor and adductor of femur; Vaughan, 1959, 1970b). M. obturator internus is absent in bats, probably as a result of hip modifications associated with 90 8 rotation of the hindlimbs (Simmons, 1994). The obturator foramen in Icaronycteris, Archaeonycteris,Hassianycteris, and Palaeochiropteryx is of the normal mammalian morphology with a well-defined rim. This is similar to the condition seen in most extant bats. In contrast, the obturator foramen is partially infilled by a thin, bony sheet along the posteroventral rim in Rhinopomatidae, Rhinolophinae, Hipposiderinae, and Thyropteridae. Such infilling is absent in both outgroups, suggesting that the condition seen in the Eocene bats is primitive. H INDLIMB Character 169: Shaft of femur straight (0); or with bend that directs distal shaft dorsally (1). Although all bats share a complex series of modifications of the hip and proximal femur that serve to rotate the hindlimbs approximately 90 8 from the typical mammalian condition (Simmons, 1994), the femoral shaft and distal femur remain relatively unmodified. As in most other mammals, the shaft of the femur is straight in Icaronycteris (frontispiece, fig. 21), Archaeonycteris (fig. 3), Hassianycteris (fig. 4), and Palaeochiropteryx (fig. 2). This condition is also seen in most extant bats. In some bats, however, the shaft of the femur is bent somewhat so that the distal end is directed more dorsally (the equivalent of a lateral bend if the femur were in the position typical of nonvolant mammals). In bats, which already have a femur that projects laterally from the hip, this bend serves to raise the knee joint even higher, well above the hip joint in some forms. A bend in the shaft of the femur is seen in Rhinolophoidea and some Phyllostomidae. The shaft of the femur is straight in both outgroups, indicating that the ‘‘straight’’ condition is primitive. Character 170: Fibula complete and well developed (0); or thin and threadlike (1); or absent or entirely unossified (2). The fibula in Icaronycteris (frontispiece) and Archaeonycteris is well developed (relatively robust) and complete (ossified from knee to ankle). In contrast, the fibula in Hassianycteris and Palaeochiropteryx is relatively much thinner, almost threadlike. Among extant bats, a complete, well developed fibula is found only in Pteropodidae, some Phyllostomidae, Mystacinidae, and Molossidae. The fibula is thin, threadlike, and often only partially ossified in Emballonuridae, Rhinopomatoidea, Megadermatidae, Rhinolophidae, some Phyllostomidae, Mormoopidae, Noctilionidae, Nataloidea, Antrozoidae, and Vespertilionidae. The fibula is absent or entirely unossified in Nycteridae. Both outgroups have a complete, well developed fibula, suggesting that the condition seen in Icaronycteris and Archaeonycteris is relatively primitive. The thin, threadlike fibula seen in Hassianycteris and Palaeochiropteryx apparently represents a derived condition. Character 171: Calcar absent (0); or present (1). The calcar is a bony and/or cartilaginous rod that extends from the ankle to support the trailing edge of the uropatagium. Presence of a calcar (and m. depressor ossis styliformis, which adducts the calcar toward the lower leg) is considered to be a synapomorphy of Chiroptera (Simmons, 1994, 1995). However, there is no evidence that a calcar was present in Icaronycteris and Archaeonycteris. A calcar is lacking in every known specimen, and our examination of the calcaneum suggests that no calcar facet is 1998 85SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS present on the calcaneum of the these forms, leading us to conclude that a calcar was absent in these taxa. In contrast, a calcar is present in Hassianycteris and Palaeochiropteryx. In both taxa, the appearance of the calcar suggests that it may have been largely cartilaginous. Among extant bats, a calcar is present in most taxa; it is absent only in some Pteropodidae, Rhinopomatoidea, and some phyllostomids. Both outgroups (and all other mammals) lack a calcar, indicating that absence of the calcar is primitive. Presence of a calcar in Hassianycteris and Palaeochiropteryx represents a derived condition. Character 172: Digits II–V of foot with three phalanges (0); or two phalanges (1). Digits II–V of the foot each have three phalanges in Icaronycteris (frontispiece, fig. 23), Archaeonycteris (fig. 3), Hassianycteris (fig. 4), and Palaeochiropteryx (fig. 2). This condition is also seen in most extant bats. In contrast, only two phalanges are present in digits II–V in Hipposiderinae, Myzopodidae, and Thyropteridae. Both outgroups (and most other mammals) have three phalanges in digits II–V of the foot, suggesting that the condition seen in the Eocene bats is primitive. COMPLETENESS One problem that plagues phylogenetic analyses that include fossil taxa is that of completeness (or lack of completeness) of available data. Numerous studies have illustrated the importance of including fossils in phylogenetic analyses, demonstrating that fossils sometimes preserve information (in the form of unique combinations of primitive and derived character states) that may be crucial to resolving relationships among extant lineages (e.g., Gauthier et al., 1988; Donoghue et al., 1989; Novacek, 1992, 1994). Fossils are also important because they frequently affect character optimizations, which in turn may affect conclusions concerning the degree of support for various monophyletic groups, taxonomic diagnoses, character independence, homoplasy, and the relative timing of various evolutionary events (Simmons, 1993a). It is also obvious that fossils must be included in any study that seeks to determine the relationships of fossils to extant groups. However, inclusion of relatively incomplete taxa—be they fossils or poorly known extant groups—can sometimes dramatically reduce the degree of phylogenetic resolution obtained by increasing the number of equally parsimonious topologies (Rowe, 1988; Simmons, 1993a, 1993c). Completeness of a taxon may be defined as the percentage of characters for which it can be scored in a given analysis (Simmons, 1993a, 1993c). In practice, paleontologists have often sought to maximize completeness by focusing largely on osteological characters, omitting from consideration many softtissue or molecular characters that cannot be scored in extinct organisms (e.g., Beard, 1993). However, inclusion of fossil OTUs in an analysis with many soft-tissue characters does not necessarily lead to decreased resolution (Gauthier et al., 1988; Novacek, 1992). In some cases, inclusion of fossils can actually increase resolution by reducing the number of equally parsimonious trees, as was the case with Novacek’s (1992) hyracoid example. When the opposite occurs and inclusion of fossils increases the number of optimal trees and decreases resolution, techniques such as Adams consensus may be used to identify relationships that remain stable in all most-parsimonious trees (Simmons, 1993a). The relative stability of topological placement of a fossil taxon in phylogenetic trees appears to depend more on the particular combination of character states that it exhibits than on its completeness. For example, in an analysis of relationships among archontan mammals, Simmons (1993a) found that one fossil taxon that was only 19% complete could be placed unambiguously relative to extant lineages, while relationships of another fossil taxon that was 53% complete could not be determined unambiguously. In our view, the benefits of including fossils and soft tissue and molecular characters in a single analysis far outweigh the possible problems (see discussion below). Nevertheless, we calculated the percent completeness for each OTU in our data set in order to provide a basis for a posteriori considerations of the effects of completeness (table 6). Extant lineages in our data set were 62.0–100% 86 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY TABLE 6 Statistics on Character Codings for Current Data Set Terminal taxon Characters scored with single state a Characters scored with two or more states b Characters scored ‘‘ 2 ’’ Characters scored ‘‘?’’ Percent completeness c Scandentia Dermoptera Pteropodidae Icaronycteris Archaeonycteris Hassianycteris Palaeochiropteryx Emballonuridae 161 (77.4%) 172 (82.6%) 176 (84.6%) 71 (34.1%) 59 (28.4%) 61 (29.3%) 75 (36.1%) 166 (79.8%) 3 (1.4%) 0 26 (12.5%) 0 0 2 (1.0%) 1 (0.5%) 26 (12.5%) 19 (9.1%) 18 (8.7%) 6 (2.9%) 3 (1.4%) 1 (0.5%) 1 (0.5%) 1 (0.5%) 6 (2.9%) 25 (12.0%) 18 (8.7%) 0 134 (64.4%) 148 (71.2%) 144 (69.2%) 131 (63.0%) 10 (4.8%) 88.0% 91.3% 100% 35.6% 28.9% 30.8% 37.1% 95.2% Rhinopomatidae Craseonycteridae Nycteridae Megadermatidae Rhinolophinae Hipposiderinae Phyllostomidae Mormoopidae Noctilionidae Mystacinidae Myzopodidae Thyropteridae Furipteridae Natalidae Antrozoidae Tomopeatinae Molossinae 185 (89.0%) 125 (60.1%) 178 (85.6%) 183 (88.0%) 185 (89.0%) 170 (81.8%) 146 (70.0%) 172 (82.7%) 192 (92.3%) 132 (63.5%) 129 (62.0%) 174 (83.7%) 147 (70.7%) 181 (87.0%) 148 (71.2%) 125 (60.1%) 182 (87.5%) 1 (0.5%) 0 6 (2.9%) 9 (4.3%) 5 (2.4%) 13 (6.2%) 57 (27.6%) 13 (6.2%) 0 0 0 1 (0.5%) 1 (0.5%) 0 1 (0.5%) 0 21 (10.1%) 5 (2.4%) 5 (2.4%) 7 (3.4%) 2 (1.0%) 5 (2.4%) 6 (2.9%) 5 (2.4%) 5 (2.4%) 7 (3.4%) 6 (2.9%) 3 (1.4%) 3 (1.4%) 3 (1.4%) 2 (1.0%) 3 (1.4%) 4 (1.9%) 2 (1.0%) 17 (8.1%) 78 (37.5%) 17 (8.1%) 14 (6.7%) 13 (6.2%) 19 (9.1%) 0 18 (8.7%) 9 (4.3%) 70 (33.6%) 76 (36.6%) 30 (13.4%) 57 (27.4%) 25 (12.0%) 56 (26.9%) 79 (38.0%) 3 (1.4%) 91.9% 62.5% 91.9% 93.3% 93.8% 90.9% 100% 91.3% 95.7% 66.4% 63.4% 85.6% 72.6% 88.0% 73.1% 62.0% 98.6% Vespertilioninae Miniopterinae Myotinae Murininae Kerivoulinae 167 (80.3%) 145 (69.7%) 187 (89.9%) 132 (63.5%) 145 (69.7%) 22 (10.6%) 0 9 (4.3%) 0 1 (0.5%) 3 (1.4%) 5 (2.4%) 2 (1.0%) 4 (1.9%) 3 (1.4%) 16 (7.7%) 58 (27.9%) 10 (4.8%) 72 (34.6%) 59 (28.4%) 92.3% 72.1% 95.2% 65.4% 71.6% Total 4471 (71.7%) 218 (3.5%) 145 (2.3%) 1406 (22.5%) 77.5% a Includes only characters states 0, 1, 2, etc.; does not include characters scored as inapplicable (‘‘ 2 ’’) or missing (‘‘?’’). Percentage reflects the percent of the total characters (208) used in the analysis. b Includes both cases of uncertainty and taxonomic polymorphism. c Completeness is defined as the percentage of characters for which a taxon can be scored based on available data; it was calculated by subtracting the percentage of characters scored ‘‘?’’ from 100%. complete; the fossils ranged from 28.9 to 37.1% complete. METHODS OF PHYLOGENETIC ANALYSIS The present study follows Simmons (1998) in adopting a character consensus (‘‘total evidence’’) approach to phylogeny reconstruction. In recent years there has been much discussion of the relative merits of character consensus versus taxonomic consensus methods for resolving systematic problems (Kluge, 1989; Barrett et al., 1991, 1993; Swofford, 1991; Bull et al., 1993; de Queiroz, 1993; Eernisse and Kluge, 1993; Kluge and Wolf, 1993; Nelson, 1993; Chippendale and Weins, 1994; Hulsenbeck et al., 1994; de Queiroz et al., 1995; Farris et al., 1995; Miyamoto and Fitch, 1995). As discussed by Simmons (1993a, 1998), character congruence is the most sensible approach to the current data set because the data consist 1998 87SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS principally of discrete-state morphological characters that cannot be reasonably partitioned. Available restriction-site data are not capable of resolving interfamilial relationships in the absence of other data (Baker et al., 1991a), hence taxonomic congruence is not a useful option. The 208 discrete characters used in the present study (see appendix 2) were scored for phylogenetic analysis, and the resulting data matrix (appendix 3) was analyzed using PAUP version 3.1.1 (Swofford, 1993). All transformations were unordered. A heuristic search with a random-addition sequence and 1000 repetitions was used to find most-parsimonious trees. Near-most-parsimonious trees (one to six steps longer) were identified in subsequent heuristic searches using the same parameters, and a decay analysis was performed following the methods of Bremer (1988). Decay values for strongly supported clades were obtained by using constrained heuristic analyses to identify the shortest trees that did not include a particular clade. A bootstrap analysis using heuristic methods (random-addition sequence, 10 repetitions for each of 1000 bootstrap replicates) was also used to evaluate the relative support for various groupings. MacClade version 3.0 (Maddison and Maddison, 1992) was used for data entry and examination of characterstate distributions. Three complete sets of phylogenetic analyses were conducted: (1) an analysis including all characters but excluding the fossil taxa; (2) an analysis including all characters and all taxa; and (3) an analysis including all taxa but excluding those characters that could not be scored in any of the fossil forms. The first analysis was designed to provide a starting point by evaluating relationships of extant lineages in the context of the revised data matrix; this effectively represents an updated version of Simmons’ (1998) analysis. Our second analysis represents the principal goal of the project, a character congruence study including both Eocene fossil genera and extant lineages. The third and final analysis was designed to evaluate the effects of soft-tissue and molecular characters on the outcome of a phylogenetic analysis including fossil forms that cannot be scored for these features. ANALYSIS OF CHARACTER TRANSFORMATIONS Character transformations were analyzed by mapping character-state distributions onto the shortest trees derived from the second set of phylogenetic analyses described above (i.e., those obtained using all characters and all taxa). Optimizations were calculated using both the ACTRAN (accelerated transformation optimization) and DELTRAN (delayed transformation optimization) options of PAUP version 3.1.1 (Swofford, 1993). MacClade version 3.0 (Maddison and Maddison, 1992) was used to visualize the results of character mapping. As discussed by Simmons (1993a), two kinds of character transformations may be recognized during the optimization process: unequivocal transformations, which have only one parsimonious placement on the optimal tree(s), and equivocal transformations, which can be parsimoniously arranged in two or more ways. Both ACTRAN and DELTRAN place unequivocal transformations on a given tree in the same manner, but they treat equivocal transformations differently. ACTRAN forces transformations to the lowest possible points on the tree, and thus favors hypotheses of reversal over hypotheses of convergence. Conversely, DELTRAN forces transformations to the highest possible points on a tree, thus favoring hypotheses of convergence over reversal. DELTRAN is often favored in studies involving fossils because it places transformations at the minimal level at which they can be observed (i.e., in the face of missing data it does not conclude that derived states exist below the level at which they can be demonstrated). In this study, we focused only on transformations that apply at and below the nodes where fossil taxa join the tree. We compared the results of ACTRAN and DELTRAN optimizations in order to identify all equally parsimonious arrangements of equivocal transformations, and interpreted our observations in the context of what is known about the function of various structures and structural complexes. Results of these analyses are presented below under ‘‘Character Evolution in Early Chiropterans.’’ 88 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY PHYLOGENETIC RESULTS Fig. 35. Results of Analysis 1, which included all characters but omitted the fossil taxa (see text for discussion). Parsimony analysis resulted in a single most-parsimonious tree (662 steps; CI 5 0.403 and RI 5 0.578), which is shown here. The numbers below internal branches indicate the percentage of bootstrap replicates in which each clade appeared; numbers above the branches are decay values (the minimum number of additional steps required to collapse each clade). The bootstrap analysis was constrained to consider only trees in which Chiroptera was monophyletic; this was done because most of the characters that support bat monophyly (table 5) were omitted from this study. RESULTS OF ANALYSES Analysis 1: All characters, fossil taxa excluded (fig. 35). As noted above, this analysis was designed to provide a starting point by evaluating relationships of extant lineages in the context of the revised data matrix. This effectively represents an updated version of Simmons’ (1998) analysis, and illustrates the effects of changes that we made in the data matrix (e.g., corrections, new characters, inclusion of hyoid data for Antrozoidae; see above). Analysis of the revised data set resulted in a single most parsimonious tree (661 steps; CI 5 0.404, RI 5 0.579) shown in figure 35. Comparisons of the results of our Analysis 1 (fig. 35) with Simmons (1998) tree (fig. 20) reveals several topological differences, all within the yangochiropteran part of the tree. First, relationships within Nataloidea are fully resolved in our Analysis 1 tree, with Thyropteridae unambiguously placed as the sister-group of the Furipteridae 1 Natalidae clade. This represents a trivial change, since this topology occurred in one of the two most parsimonious trees found by Simmons (1998) and received higher bootstrap support than did any of the alternatives in that study. A potentially more significant change in our Analysis 1 tree is placement of Antrozoidae 1998 89SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS as the sister-taxon of Vespertilionidae (i.e., Vespertilioninae 1 Miniopterinae 1 Myotinae 1 Murininae 1 Kerivoulinae) rather than as the sister-taxon of Molossidae. This suggests that Molossoidea as defined by Simmons (1998) may not be monophyletic, and that Antrozoidae may form a clade with vespertilionids as traditionally thought (e.g., by Koopman, 1993, 1994). However, support for the Antrozoidae 1 Vespertilionidae clade was very low in Analysis 1 (bootstrap value 5 12%; minimum of one additional step to collapse clade), indicating considerable uncertainty regarding this grouping. The same is also true of other clades that represent changes from Simmons (1998) topology: (1) placement of Molossidae (Tomopeatinae 1 Molossinae) as the sister-group of the Antrozoidae 1 Vespertilionidae clade (bootstrap value 5 33%; minimum of one additional step to collapse clade); and (2) placement of Mystacinidae as the sister-taxon of Molossidae 1 Antrozoidae 1 Vespertilionidae (bootstrap value 5 20%; minimum of one additional step to collapse clade). In essence, all of these changes represent rearrangements of relationships at nodes that were poorly supported in Simmons (1998) analysis and remain poorly supported in our analyses of the updated data set. Analysis 2: All characters, all taxa (fig. 36). This analysis represents the principal goal of our project, a character-congruence study including both Eocene fossil genera and extant lineages. Encouragingly, parsimony analyses of the complete data set in Analysis 2 produced a single most-parsimonious tree (fig. 36; 681 steps; CI 5 0.392, RI 5 0.587). Relationships of Icaronycteris,Archaeonycteris,Hassianycteris, and Palaeochiropteryx were fully resolved, and placement of each was moderately to strongly supported in the bootstrap and decay analyses (bootstrap values 5 54–93%; minimum of one to six additional steps to collapse clade). These levels of support are comparable to those found for many extant clades in Analysis 2. Indeed, bootstrap values for branches associated with many extant taxa (e.g., Mystacinidae, Vespertilioninae, Miniopterinae, Myotinae, Murininae, Kerivoulinae) are much lower than those associated with the fossil branches despite the fact that the extant lineages are represented by two to three times more data (table 6). The absence of soft-tissue and molecular character data clearly does not preclude relatively secure placement of the fossil forms in this study. Results of our analysis indicate that Icaronycteris,Archaeonycteris,Hassianycteris, and Palaeochiropteryx represent a series of consecutive sister-taxa to the microchiropteran crown group (i.e., the group comprised of all extant lineages). They do not form a paraphyletic group ancestral to both Megachiroptera and Microchiroptera (e.g., Eochiroptera sensu Van Valen, 1979), a monophyletic group within Microchiroptera (e.g., Palaeochiropterygoidea sensu Smith, 1977), or have special affinities with various extant microchiropteran superfamilies (as suggested by Smith and Storch, 1981). Of the four Eocene genera, Palaeochiropteryx shares the most derived traits with extant microchiropterans. Hassianycteris and Archaeonycteris are consecutive sister-taxa to the clade including Palaeochiropteryx and the microchiropteran crown group, and Icaronycteris occupies the basalmost branch in the microchiropteran part of the tree. Comparisons of the results of Analyses 1 and 2 (figs. 35, 36) demonstrate that inclusion of Icaronycteris,Archaeonycteris,Hassianycteris, and Palaeochiropteryx in the analysis produced minor changes in topology of the tree, with the changes again centered on those parts of the tree that were weakly supported in previous analyses. Specifically, inclusion of the fossils changed the position of Antrozoidae (which in Analysis 2 now forms a clade with Molossidae, supporting monophyly of Molossoidea) and Mystacinidae (which now appears as the sister-taxon of Nataloidea 1 Molossoidea 1 Vespertilionoidea). Interestingly, all of these relationships appeared in Simmons’ (1998) tree (fig. 20), although not in the results of our Analysis 1 (fig. 35). Inclusion of the fossils affected the perceived support for many clades even when there was no effect on tree topology. This effect was most noticeable near the base of tree. Monophyly of the microchiropteran crown group (Emballonuridae 1 Yinochiroptera 1 Yangochiroptera) received only moderate support in Analysis 2 (bootstrap value 90 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY Fig. 36. Results of Analysis 2, which included all characters and all taxa (see text for discussion). Parsimony analysis resulted in a single most-parsimonious tree (680 steps; CI 5 0.393; RI 5 0.587), which is shown here. The numbers below internal branches are bootstrap values; numbers above the branches are decay values. The bootstrap analysis was constrained to consider only trees in which Chiroptera was monophyletic. 5 79%; minimum of two additional step to collapse clade), whereas it received extremely high support in Analysis 1 (bootstrap value 5 100%; minimum of 16 additional steps to collapse clade). Clearly, many of the derived traits that diagnose extant Microchiroptera evolved in a sequential pattern over time; inclusion of the fossil taxa ‘‘spreads out’’ these synapomorphies over a larger part of the tree, thus reducing perceived support for any single branch. Nevertheless, monophyly of the microchiroperan crown group remains strongly supported in contrast to other hypotheses. Alternative topologies, including those suggested by Smith and Storch (1981), appear unlikely given results of the bootstrap analysis. For example, Hassianycteris formed a clade with Yangochiroptera in only 6% of the bootstrap replicates. Palaeochiropteryx grouped with Yangochiroptera in only 5% of the bootstrap replicates, and formed a clade with some or all Vespertilionoidea in fewer than 1% of the bootstrap replicates. As in Simmons (1998) and Analysis 1 (fig. 20, 35), results of Analysis 2 placed Emballonuridae as the sister-group of the clade containing all other extant microchiropteran lin- 1998 97SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS TABLE 7 A Revised Summary of Morphological Synapomorphies of Chiroptera 1) Deciduous dentition does not resemble adult dentition; deciduous teeth with long, sharp, recurved cusps 2) Palatal process of premaxilla reduced; left and right incisive foramina fused in midsaggital plane 3) Postpalatine torus absent 4) Jugal reduced and jugolacrimal contact lost 5) Two entotympanic elements in the floor of the middle-ear cavity: a large caudal element and a small rostral element associated with the internal carotid artery 6) Tegmen tympani tapers to an elongate process that projects into the middle-ear cavity medial to the epitympanic recess 7) Proximal stapedial artery enters cranial cavity medial to the tegmen tympani; ramus inferior passes anteriorly dorsal to the tegmen tympani 8) Enlarged fenestra rotundum 9) Vomeronasal epithelial tube absent 10) Accessory olfactory bulb absent 11) Posterior laminae present on ribs 12) Modification of scapula: reorientation of scapular spine and modification of shape of scapular fossae; reduction in of height of spine; presence of a well-developed transverse scapular ligament; presence of at least two facets in infraspinous fossa 13) Modification of elbow: reduction of olecranon process and humeral articular surface on ulna; presence of ulnar patella; absence of olecranon fossa on humerus 14) Absence of supinator ridge on humerus 15) Absence of entepicondylar foramen in humerus 16) Occipitopollicalis muscle and cephalic vein present in leading edge of propatagium 17) Digits II–V of forelimb elongated with complex carpometacarpal and intermetacarpal joints, support enlarged interdigital flight membranes (patagia); digits III–V lack claws 18) Modification of hip joint: 90 8 rotation of hindlimbs effected by reorientation of acetabulum and shaft of femur; neck of femur reduced; ischium tilted dorsolaterally; anterior pubes widely flared and pubic spine present; absence of m. obturator internus 19) Sacrum terminates posterior to midpoint of acetabulum 20) Absence of m. gluteus minimus 21) Absence of m. sartorius 22) Vastus muscle complex not differentiated 23) Modification of ankle joint: reorientation of upper ankle joint facets on calcaneum and astragalus; trochlea of astragalus convex, lacks medial and lateral guiding ridges; tuber of calcaneum projects in plantolateral direction away from ankle and foot; peroneal process absent; sustentacular process of calcaneum reduced, calcaneoastragalar and sustentacular facets on calcaneum and astragalus coalesced; absence of groove on astragalus for tendon of m. flexor digitorum fibularis 24) Entocuneiform proximodistally shortened, with flat, triangular distal facet 25) Elongation of proximal phalanx of digit I of foot 26) Embryonic disc oriented toward tubo–uterine junction at time of implantation 27) Differentiation of a free, glandlike yolk sac 28) Preplacenta and early chorioallantoic placenta diffuse or horseshoe-shaped, with definitive placenta reduced to a more localized discoidal structure 29) Definitive chorioallantoic placenta endotheliochorial 30) Baculum present 31) Left central lobe of liver separate from other lobes or partially fused with right central lobe 32) Caecum absent 33) Cortical somatosensory representation of forelimb reverse of that in other mammals teris,Hassianycteris,Palaeochiropteryx, and most members of the microchiropteran crown group. Because the type of periotic attachment could not be scored in available specimens of Icaronycteris, we cannot be sure if this character transformation occurred before or after differentiation of Icaronycteris from the stock leading to extant microchiropterans. The next highest node in the microchirop- 98 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY teran tree unites Archaeonycteris with Hassianycteris,Palaeochiropteryx, and the microchiropteran crown group. As noted above, loosening of the connection between the periotic and basisphenoid (character 25) minimally applies at this level in the tree. Other derived characters that diagnose this clade include (1) reduction of the number of roots on P3 to two or fewer (character 19; P3 with three roots in both outgroups, Pteropodidae, and Icaronycteris); (2) presence of a ventral accessory process on cervical vertebra 4 (character 76; accessory process absent from C4 in both outgroups, Pteropodidae, and Icaronycteris); and (3) absence of ossified third phalanx in wing digit III (character 152; fully ossified third phalanx present in both outgroups and Icaronycteris). Of these characters, reduction of the number of roots on P3 and presence of a ventral accessory process on C4 are synapomorphies that unequivocally diagnose the clade comprising Archaeonycteris and the lineage leading to the microchiropteran crown group. The absence of an ossified third phalanx on wing digit III represents an equivocal transformation because this condition also characterizes Pteropodidae. Alternative explanations for the observed pattern include (1) independent loss of ossification of this phalanx in pteropodids and the lineage leading to extant microchiropterans; or (2) loss of ossification of the third phalanx in the common ancestor of bats, and secondary acquisition of ossification of this element in Icaronycteris. Both hypotheses are equally parsimonious, and they also seem equally likely given the complex pattern seen elsewhere in the tree (e.g., three patterns occur within Yangochiroptera—full ossification [secondarily acquired once], ossification only at phalanx base [acquired twice], and no ossification [secondarily acquired once]). Other character transformations that occur somewhere near the base of the microchiropteran tree (we cannot determine exactly where because of missing data) involve evolution of the phanerocochlear condition (character 27), a deep, constricted stapedial fossa (character 31), presence of anterior laminae on the ribs (character 86), broad posterior laminae on the ribs (character 89), and presence of a dorsal articular facet on the scapula (character 100). The cryptocochlear condition is seen in both outgroups and Pteropodidae; the phanerocochlear condition occurs in Hassianycteris,Palaeochiropteryx, and primitively in the microchiropteran crown group. The stapedial fossa in the outgroups and Pteropodidae is indistinct or shallow and broad; this contrasts sharply with the deep, constricted stapedial fossae seen in Palaeochiropteryx and primitively in the microchiropteran crown group. Because external cochlear morphology and form of the stapedial fossa could not be scored in available specimens of either Icaronycteris or Archaeonycteris, we cannot be sure where on the tree between Pteropodidae and Palaeochiropteryx the transformations to a phanerocochlear condition and a deep, constricted stapedial fossa occurred. Nor can we tell if these changes were coincident or occurred at different points in the tree. Interpretive difficulties similarly occur with other character transformations as a result of missing data. Anterior laminae are absent from the ribs of both outgroups, some Pteropodidae, and Icaronycteris, while anterior laminae are present in Palaeochiropteryx and primitively within the microchiropteran crown group. Posterior laminae are narrow in Pteropodidae and Icaronycteris, but are broad in Hassianycteris, Palaeochiropteryx, and primitively within the microchiropteran crown group. A dorsal articular facet is absent from the scapula in the outgroups, Pteropodidae, and Icaronycteris; however, it is present in Hassianycteris, Palaeochiropteryx, and primitively in the microchiropteran crown group. Because the conditions of these characters in Archaeonycteris could not be determined, we cannot be sure if the transformations in these structures occurred before or after differentiation of Archaeonycteris from the lineage leading to the microchiropteran crown group. Moving up the microchiropteran part of the tree, the next node unites Hassianycteris, Palaeochiropteryx, and the microchiropteran crown group. As noted above, presence of the phanerocochlear condition (character 27), a deep, constricted stapedial fossa (character 31), anterior laminae on the ribs (character 86), broad posterior laminae on the ribs (character 89), and a dorsal articular facet on the scapula (character 100) may apply at this 1998 99SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS level in the tree. Derived characters that unambiguously diagnose the clade comprising Hassianycteris,Palaeochiropteryx, and the microchiropteran crown group include: (1) nyctalodonty (character 21; primitive tribosphenic condition in Scandentia, Icaronycteris, and Archaeonycteris); (2) a greatly enlarged cochlea (character 26; either not enlarged or only moderately enlarged in the outgroups, Pteropodidae, Icaronycteris and Archaeonycteris); (3) distal tip of ventral process of manubrium laterally compressed (character 91; tip blunt or rounded in Dermoptera, Icaronycteris, and Archaeonycteris); (4) absence of an ossified third phalanx (claw) on wing digit II (index finger; character 151; ossified claw present in both outgroups, Pteropodidae, Icaronycteris, and Archaeonycteris); (5) fibula thin and threadlike (character 170; fibula well developed in both outgroups, Pteropodidae, Icaronycteris, and Archaeonycteris); (6) calcar present (character 171; calcar absent in both outgroups, Icaronycteris, and Archaeonycteris). Optimization of the latter character (fig. 38) is somewhat surprising given previous hypotheses (e.g., Simmons, 1994, 1995; see table 5) that presence of a calcar is a synapomorphy of Chiroptera. Our analysis suggests otherwise. A calcar seems to have been absent in the most recent common ancestor of Megachiroptera and Microchiroptera, and apparently evolved independently in Pteropodidae and in the lineage leading to extant microchiropterans. The last clade directly involving the fossil taxa analyzed in this study is that comprising Palaeochiropteryx plus the microchiropteran crown group. Only a single derived character unequivocally diagnoses this clade: presence of a ventral accessory process on cervical vertebra 5 (character 77; absent in both outgroups, Pteropodidae, Icaronycteris,Archaeonycteris, and Hassianycteris). Two other characters minimially apply at this level, but lack of data for Hassianycteris and Archaeonycteris make it impossible to determine the point of transformation on the tree. These characters include (1) a deep, constricted stapedial fossa (character 31; fossa indistinct in both outgroups, shallow and broad in Pteropodidae), and (2) presence of anterior laminae on the ribs (character 86; laminae absent in both outgroups, some Pterododidae, and Icaronycteris). FEATURES DIAGNOSING THE MICROCHIROPTERAN CROWN GROUP Character transformations that appear to diagnose the microchiropteran crown group fall into two broad categories: (1) features that can be positively attributed to this particular node (by virtue of having been scored with a different state in Hassianycteris and more distal outgroups); and (2) derived features that minimally diagnose the crown group but could not be scored in the fossils, leaving open the possibility that they may have evolved earlier (closer to the basal node) in the tree. Transformations of the latter sort may be unambiguous synapomorphies, but it is not clear at what level they apply. In the first category, we find that only three transformations in hard-tissue characters appear to diagnose the microchiropteran crown group: (1) free premaxilla (character 9; sutured premaxilla in both outgroups, Pteropodidae, and all four Eocene genera); (2) reduction to two lower premolars on each side of the jaw (character 20; three lower premolars in Pteropodidae and all four Eocene genera); and (3) a xiphisternum with prominent median keel (character 95; keel absent in both outgroups, Pteropodidae, Icaronycteris,Archaeonycteris, and Palaeochiropteryx). In comparison to the lists of features diagnosing more inclusive clades (see above), these synapomorphies by themselves are not compelling. For example, few workers would agree that a free premaxilla is likely primitive for all extant microchiropterans (as suggested by optimization of this character on our tree; fig. 39) because yangochiropteran bats lack this complex specialization. Instead, yangochiropterans have a premaxilla that is firmly fused to the maxilla in the adult. The free premaxilla is a feature unique among mammals, and we agree that it is hard to imagine either its loss or independent origin in two different groups (which are, of course, the two most-parsimonious hypotheses given the topology of our optimal tree). Reduction from three to two lower premolars 100 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY Fig. 39. Premaxilla–maxilla articulations (character 9) mapped on the phylogeny from figure 36. Presence of a free premaxilla appears to be a synapomorphy of the microchiropteran crown group (although see text for discussion). Fusion of the premaxilla to the maxilla is a synapomorphy of Yangochiroptera. The condition for Pteropodidae is depicted as ‘‘uncertain’’ because two conditions occur within this group: the premaxilla is sutured to the maxilla in most pteropodids, but these elements are fused in other taxa. This optimization suggests that a suture joint between the premaxilla and maxilla is primitive for Pteropodidae. on each side is a character that is similarly troublesome, because presence of three lower premolars characterizes many groups of extant microchiropterans that nest well up within the crown group (e.g., Rhinolophidae, Mormoopidae, Nataloidea, many vespertilionids). This is also true of the xiphisternal keel. Although presence of a keel appears to diagnose the microchiropteran crown group, many lineages within that clade lack a keel on the xiphisternum (e.g., Rhinopomatidae, Hipposiderinae, some Phyllostomidae, Noctilionidae, Myzopodidae, Molossoidea, some Vespertilioninae). Despite the relative weakness of these data, monophyly of the microchiropteran crown group is indirectly supported by a broad range of other characters that unfortunately have somewhat ambiguous distributions. These transformations minimally diagnose the microchiropteran crown group, but might have evolved earlier in the tree; we could not score them in fossil sister-taxa of the microchiropteran crown group. Such features are numerous and include the following: (1) presence of a tragus (character 2; absent in outgroups and Pteropodidae); (2) aquaeductus cochleae small or absent (character 33; large in Dermoptera and Pteropodidae); (3) presence of sophisticated echolocation (character 36; absent in outgroups and Pteropodidae); (4) m. styloglossus originates from ventral surface of midpoint of stylohyal (character 58; originates from expanded tip of stylohyal and/or adjacent surface of skull in outgroups and Pteropodidae); (5) clavicle articulates with coracoid process of scapula (character 113; clavicle articulates with acromion in both outgroups, Pteropodidae, and Icaronycteris); (6) m. spinotrapezius clearly differentiated from trapezius complex (character 124; m. spinotrapezius not differentiated from trapezius complex in both outgroups and Pteropodidae); (7) origin of m. acromiodeltoideus does not include thoracic vertebra 6 (character 127; origin does include T6 in both outgroups and Pteropodidae); (8) spinal cord with angle between dorsal horns 1998 101SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS of 0–25 8 (character 190; angle between dorsal horns 35–50 8 in both outgroups, 70–80 8 in Pteropodidae); and (9) inferior colliculus larger than superior colliculus (character 191; inferior colliculus significantly smaller than superior colliculus in both outgroups and Pteropodidae). One other transformation may diagnose the microchiropteran crown group, although it has been documented only in Yinochiroptera and Yangochiroptera (no data for Emballonuridae): m. flexor digitorum profundus does not insert on digit II of wing (character 154; muscle does insert on digit II in both outgroups and Pteropodidae). Even if these features evolved in a stepwise fashion up the tree, at least a few probably represent true synapomorphies of the microchiropteran crown group. CHARACTER TRANSFORMATIONS AT BASAL NODES: A FUNCTIONAL PERSPECTIVE In the previous section, we described the character transformations that seem to apply at each node in the basal part of the microchiropteran tree. An interesting pattern also emerges when these transformations are viewed from a functional perspective. Rather than changes in each organ system being concentrated at one or two nodes, we see a pattern of stepwise changes in multiple functional systems as we move up the microchiropteran tree. This suggests a complex, mosaic pattern of evolution in which several organ systems were being refined simultaneously. Overall, the changes associated with the facial region and masticatory apparatus are relatively minor compared to those seen in other systems. In order of their appearance along the backbone of the microchiropteran tree (beginning at the base, prior to divergence of Icaronycteris), these transformations include (1) evolution of an elongate angular process (though this may ultimately prove to be plesiomorphic for bats), (2) reduction of the number of roots on P3, (3) evolution of nyctalodonty, and (4) reduction in the number of lower premolars and modification of the simple sutured connection between the premaxilla and maxilla. Functional implications of these changes remain obscure, although Slaughter (1970) noted that reduction and simplification of the premolar dentition in early bats appears correlated with shortening of the face (brachycephaly). It is possible that these trends may have been related to the need to focus the ears anteriorly and reduce interference with returning echolocation calls—the auditory equivalent of the type of facial shortening generally associated with evolution of binocular vision. Another possibility is that loss of premolars and reduction in the number of roots on the remaining teeth represent mechanisms for mass reduction. Flying animals must generate adequate lift to remain airborne, and lift requirements increase with increasing body mass (Norberg, 1985, 1986a, 1987, 1990; Rayner, 1986; Norberg and Rayner, 1987). Distribution of body mass is also important because it affects the position of the center of mass ( 5 center of gravity), which in turn influences flight efficiency. The head is the heaviest part of the body in most vertebrates in part because of the density of the teeth. In addition to the effect that a heavy head may have on total body mass and the location of the center of mass, head mass also affect the size of neck muscles needed to support the head and resist torque (Bu¨hler, 1992). Dental reduction in birds (most of which lack teeth entirely) is widely regarded as a specialization that increased flight efficiency because it simultaneously reduced total mass and concentrated more of the body mass near the center of gravity between the wings (e.g., Welty, 1955; Stahl, 1985; Bu¨hler, 1992; Feduccia, 1996). Dental reduction in pterosaurs has also been noted as a possible adaptation for reducing body mass and increasing flight efficiency (e.g., Stahl, 1985; Bu¨hler, 1992). Because bats are mammals that rely on their dentition for food processing, extreme dental reduction is rare (limitied mostly to nectarivorous taxa). However, even small reductions in mass may contribute to flight efficency. The free premaxilla present in Emballonuridae and Yinochiroptera (and perhaps in the most recent common ancestor of extant microchiropterans; fig. 39) provides an unusual degree of mobility in the snout. Utility of this feature has not been adequately investigated, but our experience in handling 102 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY live emballonurids suggests that dorsiflexion of the snout is under voluntary control. Dorsiflexion of the snout effectively increases the gape of the mouth, which may be important in feeding and/or emission of echolocation calls. With respect to the latter, it is interesting to note that the mode of emission of echolocation calls—either through the mouth (oral emission) or through the nasal passages (nasal emission)—varies among bats that have a free premaxilla. Emballonuridae, Craseonycteridae, and perhaps Rhinopomatidae are oral emitters, while Rhinolophoidea and perhaps Rhinopomaidae are nasal emitters (Pederson, 1993). Given topology of our tree (fig. 36), it appears that oral emission is primitive and nasal emission derived as suggested by Van Valen (1979) and Pederson (1993). The extent to which dorsiflexion of the snout may have played a role in the evolution of nasal emission in yinochiropteran bats has not been explored. Free movement of the premaxilla may also be important in prey manipulation, particularly in those taxa that take large arthropod or small vertebrate prey. Fenton (personal commun.) observed a Nycteris grandis eating a Nycteris thebiaca, and noted that the former used its upper lips extensively and manipulated its prey in an almost closedmouth fashion. The extent to which a free premaxilla may facilitate the capacity to handle large prey items has yet to be investigated. Changes in the basicranium and ear region also appear to have evolved in a stepwise fashion. The first modifications to appear (prior to the divergence of Icaronycteris from the microchiropteran stem stock) included a moderately enlarged cochlea, enlarged orbicular apophysis on the malleus, and an expanded cranial tip on the stylohyal. All of these features are probably related to the evolution of echolocation. As noted earlier, presence of a large orbicular apophysis may improve the ability of the middle ear ossicles to transmit high-frequency sounds with a minimum time delay, or may play a role in the avoidance of self-deafening (see discussion under character 35 above). The fact that all known extant microchiropterans have an enlarged orbicular apophysis—and all use echolocation—leads us to suspect that whatever the function(s) of this structure, an enlarged orbicular apophysis evolved to facilitate efficient function of the ear in echolocation. Likewise, the expanded tip on the stylohyal may represent a refinement of the system for producing echolocation calls. The microchiropteran echolocation system depends upon calls produced in the larynx, which is supported in part by muscles associated with the hyoid apparatus. The expanded tip on the stylohyal—which is firmly attached to the bulla—serves to anchor the hyoid apparatus to the skull, thus providing a stable attachment site for muscles of the throat. This attachment may be important in supporting the larynx during production of echolocation calls. Enlargement of the cochlea (specifically the basal turn) in microchiropterans appears to be a specialization for increased sensitivity to high-frequency sounds ( . 20 kHz), such as the returning echoes from vocalizations used in echolocation (Henson, 1970; Novacek, 1985a, 1987, 1991; Habersetzer and Storch, 1992). In all mammals, the basal turn of the cochlea is the region where high-frequency sounds are perceived (Henson, 1970; Dallos, 1973; Bruns, 1979; Bruns et al., 1983–1984; Harris and Dallos, 1984). Enlargement of this region, which effectively lengthens the basal portion of the basilar membrane, apparently increases sensitivity to high-frequency sounds and slight frequency shifts in this range (Henson, 1970; Dallos, 1973; Bruns, 1979; Bruns et al., 1983–1984; Harris and Dallos, 1984). In the lineage leading to extant Microchiroptera, cochlear enlargement seems to have evolved in a relatively continuous fashion that we chose to score as a series of steps (fig. 40). The primitive condition for bats was the presence of an unenlarged cochlea, one comparable in size to those seen in other mammals of similar body size. A moderately enlarged cochlea evolved prior to the divergence of Icaronycteris, to be followed subsequently by even greater enlargement before the divergence of Hassianycteris. Moderate enlargement of the cochlea does not by itself indicate that a bat could echolocate, because some nonecholocating pteropodids have a moderately enlarged cochlea (see discussion under character 26 above). However, the 1998 103SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS Fig. 40. Cochlear enlargement (character 26) mapped on the phylogeny from figure 36. In the lineage leading to extant microchiropterans, a ‘‘moderately enlarged’’ cochlea evolved prior to the divergence of Icaronycteris; a ‘‘greatly enlarged’’ cochlea evolved prior to the divergence of Hassianycteris. Reversals to the former condition apparently occurred independently in Mystacinidae and within Megadermatidae and Phyllostomidae (note that the latter taxa are marked as ‘‘uncertain’’ as a result of taxonomic polymorphism). This optimization also suggests that a moderately enlarged cochlea evolved independently within Pteropodidae (see text for discussion). combination of a moderately enlarged cochlea, enlarged orbicular apophysis on the malleus, and an expanded cranial tip on the stylohyal are seen in extant bats only in forms that use echolocation. We therefore follow previous authors (e.g., Novacek, 1985a, 1987, 1991; Habersetzer and Storch, 1992) in inferring that Icaronycteris,Archaeonycteris,Hassianycteris, and Palaeochiropteryx used echolocation. We agree with Pye (1968: 797), who observed that The Eocene brought mammals mean And bats began to sing; Their food they found by ultrasound And chased it on the wing. The implications of cochlear morphology for reconstructing foraging strategies is discussed in depth below under ‘‘Foraging ecology of Eocene bats.’’ Loose attachment of the periotic to the basicranium evolved prior to the divergence of Archaeonycteris, although we cannot determine the exact level in the phylogeny because this character could not be scored in Icaronycteris. Based on the presumed function of this loose attachment—cochlear isolation, thought to function in reducing bone conduction of laryngeal vibrations (Henson, 1970; Van Valen, 1979)—we surmise that this feature evolved in concert with the early stages of cochlear enlargement and facilitated the evolution of echolocation. Another transformation that took place in the microchiropteran lineage prior to the divergence of Hassianycteris was evolution of a phanerocochlear cochlea. Unfortunately, we cannot determine the level of this transformation in the phylogenetic tree because this character could not be scored in Icaronycteris or Archaeonycteris. The phanerocochlear condition was explained by Novacek (1991: 84) as ‘‘an accommodation to the problem of ‘‘packing’’ middle ear structures in a space constrained by the expanded cochlea.’’ As discussed above under character 27, this feature is probably linked to cochlear expansion, although we cannot be sure at what point the phanerocochlear condition evolved relative to changes in cochlear size. 104 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY Prior to the divergence of Palaeochiropteryx, yet another transformation took place: evolution of a deep, constricted fossa for m. stapedius. Again, we cannot determine the level of this transformation in the phylogenetic tree because this character could not be scored in Icaronycteris,Archaeonycteris,or Hassianycteris. The function of this transformation remains somewhat obscure, although it is possible that it is in some way linked to the system for avoiding self-deafening that is used by many echolocating bats (Henson, 1964, 1965, 1966, 1967a, 1979; Jen and Suga, 1976; Fenton et al., 1995; see discussion below). Transformations in features of the postcranial skeleton also seem to have evolved in a series of steps up the tree. The first change in the postcranium was proximal extension of the trochiter up to the level of the head of the humerus, which occurred prior to the divergence of Icaronycteris. Several other derived traits evolved subsequently, although we cannot be sure at what level because they could not be scored in Archaeonycteris. These traits include (1) presence of a dorsal articular facet on the scapula, (2) presence of anterior laminae on the ribs, and (3) presence of broad posterior laminae on the ribs. Transformations that occurred subsequent to the divergence of Archaeonycteris but prior to the divergence of Hassianycteris include (1) laterally compressed ventral process of the manubrium, (2) loss of an ossified third phalanx in digit II of the wing, (3) reduction of fibula to a thin, threadlike element, and (4) presence of a calcar. Dividing these features by anatomical region, we see progressive changes occurring in different functional units. Modifications of the shoulder region began with enlargement of the trochiter, which preceded evolution of a dorsal articular facet on the scapula. A functional shoulder-locking mechanism, which requires a secondary articulation between the trochiter and dorsal articular facet, was clearly present by the time that Hassianycteris diverged from the lineage leading to extant microchiropterans. In terms of morphology, the primitive shoulder-lock apparently consisted of a trochiter that extended up to but not beyond the humeral head, and a dorsal articular facet in the form of a small groove or oval situated on the anteromedial rim of the glenoid fossa. In the forelimb skeleton, we see progressive reduction in the number of phalanges in digit II of the wing (the index finger) beginning after the onset of evolutionary changes in the shoulder region (fig. 41). Because the thumb is relatively small in all bats (at least in comparison to the other digits of the hand), digit II forms the leading edge of the dactylopatagium (‘‘hand wing’’) near the wrist. The dactylopatagium plays an important role in both lift generation and steering during flight (Vaughan, 1959; Norberg, 1969, 1970, 1972b, 1976; Hill and Smith, 1984). Stiffness of leading edge of the dactylopatagium—particularly the section known as the dactylopatagium minus, which extends between the second and third digits—is critical to the ability of the wing to resist bending and twisting forces (Norberg, 1969, 1970, 1972b). This functional unit is also important in maintaining wing camber (Vaughan, 1959, 1970c; Norberg, 1970, 1972b, 1976). In bats lacking the distal phalanges on digit II, a ligamentous connection runs between the end of first phalanx of digit II and the base of the second phalanx in digit III (Norberg, 1969, 1970, 1972b). This ligament is kept under continuous tension by the structure of digit III, which is bent posterodorsally due to the structure of the metacarpophalangeal joint (Norberg, 1969, 1970, 1972b). This arrangement results in a convex frame within which the dactylopatagium minus is kept stretched under tension, thus forming a unit that is stiff in the plane of the membrane (Norberg, 1969, 1970, 1972b). Photographs of flying bats show that the airstream produces little deformation of the dactylopatagium minus or its supporting elements during flight (Norberg, 1969, 1970). The tensile strength of the second digit is critical to the funtion of this unit, since the second digit must resist the forces placed upon the dactylopatagium minus during flight. Reduction of the number of phalanges in digit II apparently took place in a sequential fashion in bats, with complete loss of ossified phalanges evolving independently at least three times in the microchiropteran crown group (fig. 41). It seems likely that loss of phalanges served 1998 105SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS Fig. 41. Changes in the number of phalanges on wing digit II (the index finger; characters 149, 150, and 151) mapped on the phylogeny from figure 36. Presence of three ossified phalanges on digit II is primitive for bats. Reduction to two ossified phalanges occurred just prior to the divergence of Hassianycteris from the lineage leading to the microchiropteran crown group. The microchiropteran crown group is diagnosed by further reduction to only one ossified phalanx. Compete phalangeal reduction (no ossified phalanges on wing digit II) apparently evolved independently at least three times in the crown group—in Emballonuridae, Nataloidea, and either in the common ancestor of Rhinolophoidea or independently in Nycteridae and Rhinolophidae. Presence of two ossified phalanges in Rhinopomatidae appears to be a reversal rather than retention of the primitive condition. to increase the tensile strength per unit mass of the distal second digit by replacing articulated phalanges (which together have a relatively low tensile strength per unit mass) with a continuous ligament that has a higher tensile strength per unit mass. Mass reduction, particularly near the distal end of the wing, contributes significantly to efficient flight performance (Swartz, 1997). Loss of the distal phalanges may represent a mechanism for distal mass reduction as constrained by the need to provide the tensile strength necessary to maintain a stiff dactylopatagium minus during flight. Modifications in the axial skeleton in the basal part of the microchiropteran tree included evolution of anterior laminae and broad posterior laminae on the ribs. Broad posterior laminae evolved prior to the divergence of Hassianycteris, although we cannot be sure of the level because we could not score this character in Archaeonycteris. Similarly, anterior laminae evolved at some point prior to the divergence of Palaeochiropteryx, although we cannot be sure of the level because we could not score this character in Archaeonycteris or Hassianycteris. In any case, these modifications are absent in Icaronycteris, so they must have occurred within the basal microchiropteran lineage. We expect that function of rib laminae is twofold: to stiffen the ribcage and to provide larger areas for muscle attachment. Muscle groups that originate directly from the ribcage include those of the m. serratus anterior complex, which is a critical component of the flight musculature. The posterior division of m. serratus anterior contributes to the downstroke, and the anterior division serves to anchor the medial edge of the scapula and may also help to initiate the upstroke of the wing (Vaughan, 1959, 1970b; Norberg, 1970, 1972a; Strickler, 1978; Hermanson and Altenbach, 1983, 1985). 106 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY A close association has been demonstrated between wingbeat and sound emission in microchiropterans under laboratory conditions (Schnitzler, 1968, 1970a, 1970b, 1971, 1973; Suthers et al., 1972; Schnitzler and Henson, 1980; Joerman and Schmidt, 1981; Heblich, 1986; Lancaster et al., 1992) and in free-flying bats foraging in nature (Kalko, 1994). Recent studies of flight, respiration, energy expenditure, and echolocation have indicated that while production of echolocation calls is extremely costly in resting bats, there is little additional cost for echolocation in flying bats (at least for search-phase calls; Kalko, 1994) because the same muscles that flap the wings also ventilate the lungs and produce the pulses of breath used to generate echolocation calls (Speakman et al., 1989; Rayner, 1991a, 1991b; Speakman and Racey, 1991; Speakman, 1993). A critical link in this system is m. serratus anterior, which ties the ribcage to the flight mechanism. The increased attachment area for this muscle complex provided by anterior and posterior laminae on the ribs—and concomitant increase in stiffness of the ribcage that we hypothesize is created by these laminae—may simultaneously increase efficiency of the flight mechanism, ventilation, and echolocation system. M. serratus anterior plays an important role in flight during the downstroke and early stages of the upstroke. A larger area of origin may provide for improved performance of this muscle complex in flight and may also facilitate ventilation of the lungs both by increasing the area of the connection between the flight apparatus and ribcage. Increased stiffness of the ribcage caused by presence of rib laminae could also faciliate exhaling and production of echolocation calls by increasing the force of elastic recoil of the ribcage upon relaxation of m. serratus anterior. We note that the echolocating bat species studied by Speakman and his colleagues (Phyllostomus hastatus,Plecotus auritus,Pipistrellus pipistrellus, and Plecotus auritus) all have some degree of development of rib laminae. It would be interesting to determine if the few microchiropterans that lack rib laminae (e.g., Molossus molossus) are as energy-efficient while flying and echolocating as are taxa with rib laminae. If rib laminae have been secondarily lost in these forms (as we infer from our phylogeny), we would expect that any loss in efficiency related to ribcage structure would have been compensated for by other mechanical changes in the flight, ventilation, and echolocation systems. Another change in the axial skeleton involved evolution of a laterally compressed ventral process on the manubrium, which appeared prior to the divergence of Hassianycteris. Changes in this structure, which forms a keel when laterally compressed, are likely to be functionally related to the flight muscles that originate from the manubrium, specifically m. pectoralis. M. pectoralis provides most of the power for the downstroke of the wings (Vaughan, 1959, 1970b; Strickler, 1978). More anteriorly located fibers in this complex (e.g., those originating from the manubrium rather than from the body of the sternum) draw the humerus downward and sharply forward, whereas the posterior fibers pull the humerus downward and backward (Vaughan, 1959). Modification of the ventral process of the manubrium may reflect changes in the relative size, moment arm, and functional importance of the anterior portion of m. pectoralis. Moving to the hindlimbs, we find the first case in which known modifications of an anatomical region seemingly evolved in a single segment of the tree rather than in a stepwise manner. In this instance, all changes occurred after the divergence of Archaeonycteris and prior to the divergence of Hassianycteris. Two transformations occurred at this level: reduction of the fibula to a thin, threadlike element, and evolution of a calcar. The functional implications of the former are not clear. Several authors have noted fibular reduction in extant microchiropterans, and have implied that reduction of the fibula is somehow associated with lack of a need for a robust fibula in bats that habitually hang under branches and do not use typical quadrupedal locomotion (Vaughan, 1959, 1970a; Walton and Walton, 1968, 1970; Howell and Pylka, 1977; Hill and Smith, 1984). This suggests that the fibula may have been reduced by default when it was no longer needed to support compressive forces associated with quadrupedial locomotion. However, the correlation between hanging behavior and a thin fibula is not per- 1998 113SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS chiropterans, in many species it is comparable to that of nocturnal muroid rodents (e.g., Rattus,Peromyscus) and marsupials (e.g., Didelphis; Suthers, 1966; Rahmann, 1967; Suthers et al., 1969; Chase, 1972; Manske and Schmidt, 1976; Bell, 1982b; Fenton, 1985; Bell and Fenton, 1986; Pettigrew, 1988; Pettigrew et al., 1988). In both Microchiroptera and Megachiroptera, visual acuity deteriorates with decreasing light levels more slowly than in humans, so bats can usually see better than humans in dim light; thresholds are similar to those reported for owls (Neuweiler, 1967; Suthers, 1970; Manske and Schmidt, 1976; Fenton, 1985; Bell and Fenton, 1986). We surmise that the earliest bats had at least the same visual capabilities as nocturnal rodents and marsupials, and may have been much better equipped (i.e., if they possessed the derived visual systems seen in primates, dermopterans, and megachiropterans; for a summary see Pettigrew et al., 1989). Reliance on the visual system alone for orientation probably precluded aerial insectivory and flight within cluttered environments (e.g., within dense vegetation and many subcanopy habitats). Nevertheless, these early bats would have benefited from many of the advantages of powered flight— rapid, energy-efficient transportation over relatively long distances and an increased foraging radius (Smith, 1977; Norberg, 1986a, 1989, 1994; Rayner, 1986; Scholey, 1986; Norberg and Rayner, 1987; Thomas, 1987), as well as easy avoidance of predators (Pomeroy, 1990). This hypothesis raises the linked questions of how and why one fully functional orientation system (vision-based orientation) would be exchanged during evolution for another (echolocation). While such a transformation is admittedly unlikely if one postulates a simple one-step process, it makes considerably more sense when viewed as a multistep process that allowed the microchiropteran lineage to successfully invade an enormous yet empty set of ecological niches for nocturnal aerial insectivores. In a rarely cited paper on the evolution of feeding strategies in bats, Gillette (1975) argued that major adaptive shifts might have evolved through what he termed ‘‘duality’’—a period during which both the primitive and derived behaviors were present simultaneously. In the case of a transformation from visual orientation to echolocation, we suggest that the early microchiropteran lineage passed through a period during which they oriented principally using vision, but also used echolocation to provide supplementary information about their surroundings. This possibility was first suggested by Pettigrew (1988: 649), although in the context of bat diphyly: I . . . think it unlikely that sonar would have been sufficiently developed in the first microbat to enable much useful guidance toward an airborne insect. The short range of ultrasound pulses makes it unlikely that it was an echo from an insect that first enticed our early microbat off its branch, as does the required neural processing which seems unlikely to have been sufficiently sophisticated on the first try. Vision seems a more likely candidate to provide the appropriate resolution and range, just as it does in many living microbats today. The primary role of sonar would then have been in the detection of obstacles, particularly in the cave roost where there is no alternative sensory channel and where three other flying vertebrates have sought shelter, independently inventing sonar on each occasion (swiftlets, oilbirds and rousette megabats). If this is correct, then the use of sonar for insect capture occurred as a modification of the avoidance system after flight was achieved. Indeed, adequate vision may have been a necessary prerequisite for the evolution of echolocation in bats. The sensory range of echolocation is relatively short (e.g., Suthers, 1970, 1978; Griffin, 1971; Fenton, 1980, 1982a, 1984, 1994a; Kick, 1982; Lawrence and Simmons, 1982), and quickly moving, flying organisms may require more information about distant obstacles than can be obtained from echolocation alone (Suthers and Wallis, 1970; Chase, 1972; Fenton, 1985; Pettigrew et al., 1988). This may be especially true in the warm, moist air of the tropics, where atmospheric attenuation of ultrasonic frequencies is greatest (Griffin, 1971; Suthers, 1978; Lawrence and Simmons, 1982). Particularly when orienting over long distances (e.g., when commuting to and from foraging areas), vision appears to be crucial for obstacle avoidance and landmark recognition (Davis, 1966; Mueller, 1966, 1968; Williams et al., 1966; Williams and Williams, 1967, 1970; Suthers, 1970). We find it hard to imagine how echolocation—which is essentially a short-range sensory system—could have evolved in a group 114 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY of flying organisms in the absence of vision for longer-range obstacle detection. Speakman (1993) argued that an early bat which relied on a combination of vision and echolocation would be a ‘‘sensory generalist’’ that might be at a selective disadvantage compared with sensory specialists that used either vision or echolocation. This hypothesis was offered as a possible explanation for the absence of echolocation in Megachiroptera, but might also apply to the earliest microchiropterans. However, we do not find this hypothesis compelling, at least as an argument against a transformation from vision to echolocation as the primary sensory system in early Microchiroptera. Early members of the microchiropteran lineage would have faced no competition from sensory specialists in echolocation, because specialized aerial echolocators—their descendants—had not yet evolved. Rather than being at a selective disadvantage when compared with vision specialists (i.e., megachiropterans), we have every reason to believe that early microchiropterans would have immediately gained several advantages when they first began to evolve echolocation, including an increased ability to detect obstacles at short range (particularly in low-light situations) and an increased ability to utilize caves as roosting places. Previous authors have suggested that the primitive echolocation system within Microchiroptera may have consisted of short broadband or multiharmonic clicks produced in the larynx (Pye, 1980; Simmons, 1980; Simmons and Stein, 1980; Fenton, 1984; Norberg, 1994; Fenton et al., 1995; Arita and Fenton, 1997). Such a system probably evolved from vocalizations used for other purposes, most likely intraspecific communication (Fenton, 1984, 1985; Novacek, 1985a). This hypothesis is supported by observations that some extant microchiropterans use echolocation calls simultaneously for orientation and communication, while others have distinct calls that are apparently used only in a social context (Mo¨hres, 1967a, 1967b; Habersetzer, 1981; Miller and Degan, 1981; Brown et al., 1983; Fenton, 1984, 1985, 1994b; Guppy et al., 1985). Megachiropteran bats use short broadband or multiharmonic FM calls in a variety of social situations, including female contact, precopulation, and hostile male–male interactions (Nelson, 1964; Fenton, 1985). Similar calls are sometimes used by microchiropteran bats, often in comparable social situations (e.g., hostile interactions between males; Porter, 1979; Fenton, 1985). Young bats of many species (both megachiropterans and microchiropterans) apparently use similar calls when they are isolated from their mothers (Fenton, 1985). Together, these observations suggest that broadband or multiharmonic clicks or buzzes were present in the vocal repertoire of the earliest bats, and thus were available as a behavioral substrate from which echolocation calls could evolve. Early members of the microchiropteran lineage, flying and using vision as well as a primitive echolocation system such as described above, would have had access to many habitats and food sources but would have been poorly equipped to fly in cluttered spaces or capture flying insects (Norberg, 1994). Perfection of echolocation for detecting, tracking, and assessing airborne prey would have increased the foraging options open to these bats, especially given the absence of competition (Fenton, 1974a, 1974b, 1980, 1982a, 1984, 1994a; Norberg, 1994; Speakman, 1995). Nocturnal flying insects offer an abundant food supply, so much so that there is little evidence that prey abundance is a limiting resource for most microchiropteran aerial insectivores, 11 at least 11 Possible exceptions to this general pattern may include aerial insectivores that summer in high latitude habitats, where the nights are very short and peak aerial insect availability occurs before it gets dark (Rydell, 1992; Speakman, 1995). It has recently been shown that some bats living under these conditions frequently fail to meet their energy requirements, instead using torpor to balance their daily energy budgets (Kunz, 1980; Kurta et al., 1987, 1989; Speakman and Racey, 1987; Audet and Fenton, 1988). Speakman (1995) suggested that it is nocturnality (and lack of adequate prey resources at night) that forces the use of torpor in these animals. Although this may be true, we suggest that it is the availability of torpor that has permitted these species to expand their summer ranges beyond the limit of most other bats. In tropical and subtropical environments (where most insectivorous microchiropteans live), nights are longer and prey densities are generally high, thus prey availability is probably not a limiting factor under normal conditions (e.g., Fenton et al., in press). 1998 115SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS those in tropical and subtropical habitats (Fenton, 1980; Fenton et al., in press). The potential advantages of evolving adaptations to exploit nocturnal flying insects as a food source are obvious. Fenton et al. (1995) noted that transformation to the sophisticated echolocation system used by extant microchiropterans to detect, track, and evaluate flying insects required production of longer, stronger (highintensity) tonal signals and concomitant evolution of a system to avoid self-deafening. Among extant bats, two such systems are known: separation of the pulse and echo in time (the ‘‘low-duty-cycle’’ approach), and separation of the pulse and echo in frequency rather than time (the ‘‘high-duty-cycle’’ approach; Fenton, 1994a, 1995; Fenton et al., 1995). Low-duty-cycle echolocation involves short signal pulses with relatively long gaps between them; high-duty-cycle echolocation involves longer pulses and shorter gaps, with pulses longer than the gaps between them (Fenton, 1994a, 1995; Fenton et al., 1995). Low-duty-cycle bats prevent self-deafening by freezing movement of the middle-ear ossicles through contraction of the middle-ear muscles during pulse emission and by reducing auditory sensitivity in the inner ear through changes in the sensory cells along the basilar membrane (Henson, 1964, 1965, 1966, 1967a, 1967b, 1970; Jen and Suga, 1976). Returning echoes are received in the gaps between pulse emissions, when the middle-ear muscles relax and auditory sensitivity is maximized. High-dutycycle bats produce long constant frequency (CF) echolocation signals that overlap with returning echoes. These bats utilize the Doppler effect, which shifts the frequency of returning echoes to a frequency different from that of the original pulses (Schnitzler, 1970b, 1973, 1987; Schuller et al., 1975; Schuller and Pollack, 1976; Schuller, 1977; Neuweiler et al., 1980; Schnitzler and Henson, 1980; Simmons and Stein, 1980; Emde and Schnitzler, 1986, 1990; Neuweiler, 1989, 1990; Grinnell, 1995). Self-deafening is reduced because the emitted pulse is dominated by frequencies outside the acoustic fovea (zone of maximum hearing sensitivity), while both the external and inner ears are sharply tuned to the frequencies of the returning echoes (Schuller and Pollack, 1976; Bruns, 1979; Neuweiler et al., 1980; Schnitzler and Henson, 1980; Vater et al., 1985; Emde and Schnitzler, 1986, 1990; Schnitzler, 1987; Vater, 1987; Neuweiler, 1990; Obrist et al., 1993; Fenton, 1994a; Fenton et al., 1995). Most extant microchiropterans are lowduty-cycle echolocators; high-duty-cycle echolocation is used only by rhinolophines, hipposiderines, and one mormoopid (Pteronotus parnellii; Novick, 1958a, 1962, 1963a, 1963b, 1977; Novick and Vaisnys, 1964; Schnitzler, 1970b, 1973, 1987; Fenton, 1974a, 1974b, 1980, 1982a, 1982b, 1984, 1990, 1994a, 1995; Schuller et al., 1975; Schuller and Pollack, 1976; Schuller, 1977; Neuweiler et al., 1980; Schnitzler and Henson, 1980; Simmons and Stein, 1980; Fenton and Bell, 1981; Neuweiler, 1984, 1989, 1990; Neuwieler and Fenton, 1988; Kalko and Schnitzler, 1989, 1993; Lancaster et al., 1992; Surlykke et al., 1993; Fenton et al., 1995; Grinnell, 1995). It seems clear that low-duty-cycle echolocation is primitive for extant Microchiroptera based on optimization of echolocation strategies on our phylogenetic tree (fig. 36). Optimization indicates that the high-duty-cycle approach evolved twice, once in the lineage leading to Rhinolophidae and once within the genus Pteronotus. This confirms previous hypotheses (e.g., Pye, 1980; Simmons, 1980; Simmons and Stein, 1980; Fenton et al., 1995) that suggested independent origins for the use of long CF signals and Doppler compensation in these two groups. The transformation from primitive to sophisticated low-duty-cycle laryngeal echolocation likely took place in several stages (Fenton, 1984), probably facilitated by the mechanical coupling of flight and ventilation discussed previously (Speakman et al., 1989; Rayner, 1991b; Speakman and Racey, 1991; Speakman, 1993). If there was indeed ‘‘no cost of echolocation for bats in flight’’ (Speakman and Racey, 1991: 421), or (more realistically) a relatively low cost, it is easy to imagine how this system might have evolved quickly. However, refinement of this system—and increased reliance on aerial insectivory—apparently brought with it an im- 116 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY portant series of evolutionary constraints, including limitations on body size and reproduction (Barclay and Brigham, 1991; Barclay, 1994, 1995; Jones, 1994; Arita and Fenton, 1997). Maximum body size may be limited in microchiropteran aerial insectivores for the several reasons, including (1) the small size of available prey (McNab, 1969; Black, 1974); (2) the metabolic requirements for sustained flight, particularly at night when thermal gliding and soaring are not possible (Rayner, 1981); (3) the mechanics of flight and aerial prey capture (Norberg, 1986a, 1994; Norberg and Rayner, 1987); (4) the effective range of echolocation calls (relatively short) and the concomitant need to be maneuverable and agile enough to catch small prey at short range (Barclay and Brigham, 1991); and (5) the coupling of flight and echolocation, which may limit the ability of large bats (which have lower wing-beat frequencies and thus lower call-repetition rates) to detect small flying prey (Jones, 1994, although see Heller, 1995). As a probable result of these constraints, microchiropteran aerial insectivores are typically very small ( , 30 g adult body weight), with only four extant species weighing more than 100 g 12 (McNab, 1969; Black, 1974; Fenton and Fleming, 1976; Krazanowski, 1977; Barclay and Brigham, 1991; Arita and Fenton, 1997). Within this size range, most larger aerial insectivores seem to be limited to relatively large insect prey (e.g., moths and large beetles; Vaughan, 1977), while smaller bats can exploit either large or small prey (Barclay and Brigham, 1991; Arita and Fenton, 1977). Returning to the problem of explaining the evolution of sophisticated echolocation— specifically our hypothesis that this form of orientation replaced a vision-based system in microchiropteran bats—we note that small body size may have precluded the retention of derived structures of the visual system (such as those seen in Primates, Dermoptera, and Megachiroptera) once neural modifications associated with echolocation began to evolve. In a review article on encephalization 12 Taphozous peli (Emballonuridae), Hipposideros commersoni (Rhinolophidae), Scotophilus nigrita (Vespertilionidae), and Cheiromeles torquatus (Molossidae). in vertebrates, Harvey and Krebs (1990: 145) noted that If overall size [of the brain] is constrained . . . it may well be that specialized enlargement of one region has to be associated with reduction in size of another ....Inother words, there may be trade-offs in the evolutionary specialization of the brain. Cooper et al. (1993a: 340) similarly concluded that . . . non-visual systems may ‘‘compete’’ with neurons of the visual system for available metabolites .... Although the use of the expression competition to describe these evolutionary events is somewhat unexplicit, the relative expansion of one cerebral structure, restrained within the confines of a braincase of limited volume, may depend upon concomitant decrease elsewhere, notwithstanding the problem of maintaining the most efficient and adaptive neuronal populations. Small body size places limitations on the size of neuron populations and may constrain the metabolic energy available for brain functions (Martin, 1981; Armstrong, 1983; Williams and Herrup, 1988; Deacon, 1990a, 1990b; Cooper et al., 1993a). Neurons have high energy requirements; the brain can consume up to 20% of circulating oxygen and glucose even though the brain constitutes only a small fraction of total body weight (Kety and Schmidt, 1948; Martin, 1981; Armstrong, 1983; Williams and Herrup, 1988; Cooper et al., 1993a). Reviewing relationships between body size, brain size, and metabolic rates in small mammals, Cooper et al. (1993a: 339) summarized their findings as follows: These arguments emphasize the conclusion that in small animals, such as moles and bats, the fraction of metabolism devoted to neurons is great and bioenergetic limits become critical. Supplying the brain with sufficient oxygen is a challenge to the body for survival. Superfluous neurons . . . are thus strongly selected against and their reduction can contribute to the animal’s fitness by improving metabolic efficiency (Ricklefs and Marks, 1984). In the early evolution of microchiropteran bats, it may not have been efficient or feasible to retain derived complex structures in the visual system (e.g., laminated dorsal lateral geniculate nucleus, large superior colliculus; Sanderson, 1986; Pettigrew et al., 1989) and simultaneously provide the neurons and metabolic energy necessary for processing increasingly complex auditory information 1998 117SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS (e.g., in a large inferior colliculus; Aitkin, 1986; Casseday and Pollak, 1988; Pollack and Casseday, 1988; Pollak and Park, 1995). From a dual orientation system employing both vision and limited echolocation, we surmise that the auditory system of microchiropteran bats simply ‘‘outcompeted’’ the visual system for the resources available. Evolutionary reduction in brain centers responsible for processing visual information could have occurred through a variety of ontogenetic changes, including truncation of development of visual centers at increasingly early stages. This hypothesis is supported by observed similarities between the relatively simple adult dorsal lateral geniculate nucleus (dLGN) in microchiropteran bats (Pettigrew et al., 1989) and early developmental stages of the dLGN in Tupaia (Casagrande and Brunso-Bechtold, 1985). If development of the dLGN in Tupaia was arrested at an early stage, the adult dLGN would resemble that of a microchiropteran bat—and eight of the derived traits defined by Pettigrew et al. (1989: 512) would be reversed as the result of a single ontogenetic change. The same would presumably hold true for the dLGN of megachiropteran bats, which is similar to that of Tupaia in most respects (Pettigrew et al., 1989). Another mechanism that might contribute to reduction in visual centers in the brain is changes in the magnitude of neuron death during early development. Neuron death is a normal part of brain development in many vertebrates, including all mammals (Finlay et al., 1987; Williams and Herrup, 1988). Not surprisingly, the magnitude of neuron death within homologous neuron populations varies phylogenetically (Finlay et al., 1987; Williams and Herrup, 1988). For example, normal development includes no death of retinal ganglion cells in fish and amphibians (Wilson, 1971; Easter et al., 1981), but 40% of retinal ganglion cells normally die in chickens (Rager, 1980), 60–70% die in rats and primates (Rakic and Riley, 1983; Crespo et al., 1985; Provis et al., 1985), and 80% die in cats (Williams et al., 1986). Neuron death is known to be a normal part of development of the dLGN in primates (Williams and Rakic, 1988) and has been shown to change the proportion of contralateral (crossed) versus ipsilateral (uncrossed) retinal projections in cats (Jacobs et al., 1984). Evolutionary changes in taxon-specific rates of neuron death in components of the visual system during development may thus affect the size and interconnections of visual centers in adults. As auditory information became increasingly important to early members of the microchiropteran lineage, we suspect that brain centers involved in processing visual information were reduced and/or simplified through developmental changes such as these. The hypothesized evolutionary reduction of the visual system in microchiropteran bats is not a unique event; similar transformations are postulated to have occurred independently in other mammalian groups that depend little on vision, including moles (Johnson, 1954; Lund and Lund, 1965, 1966; Suthers and Bradford, 1980; Kudo et al., 1988, 1991), mole rats (Bronchti et al., 1991; Cooper et al., 1993a, 1993b), and odontocete whales (Jacobs et al., 1975; Sanderson, 1986; Deacon, 1990b). In this context, it is interesting to note that odontocetes are the only other group of mammals known to use sophisticated echolocation comparable to that of microchiropterans (e.g., Norris et al., 1961; Norris, 1968; Purves and Pilleri, 1983; Pilleri, 1983, 1990). Extensive reversals in the visual system are also commonly accepted as explanations for the morphology of blind snakes, gymnophionan amphibians, and blind cave populations of fishes (e.g., Wilkens, 1971; Halpern, 1973; Repe´rant et al., 1987; Voneida and Sligar, 1976; Clairambault et al., 1980; Fritzsch et al., 1985; Himstedt and Manteuffel, 1985). Similar evolutionary reductions have taken place in other sensory systems in some mammals, such as reduction of the olfactory system in whales (Oelschla¨ger and Buhl, 1985; Oelschla¨ger, 1989) and the vomeronasal system in catarhine primates (Meredith, 1991). As noted above, these sorts of changes appear to represent modifications for efficient use of cellular and metabolic resources. When a system is no longer of critical importance, regression or reduction—-which may be interpreted as reversal in a phylogenetic context—takes place to avoid wasting cells and metabolic energy that may be better spent elsewhere (Cooper et al., 1993a). The evolution of flight and echolocation 118 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY required modifications of many anatomical and behavioral systems, and this process undoubtedly took place in several stages. To summarize, we hypothesize that flight evolved first, prior to the divergence of Megachiroptera and Microchiroptera. The first bats most likely used vision for orientation and obstacle detection in their arboreal/aerial environment. The evolution of flight was later followed by the origin of low-duty-cycle echolocation in basal members of the microchiropteran lineage. This system, which was probably derived from vocalizations originally used for intraspecific communication, was most likely simple at first, permitting orientation and detection of obstacles but not detection or tracking of airborne prey. However, due to the mechanical coupling of ventilation and flight, energy costs of echolocation to flying bats were low, and the benefits of aerial insectivory quickly led to development of a more sophisticated low-duty-cycle echolocation system capable of detecting, tracking, and assessing airborne prey. The need for an increasingly derived auditory system, combined with limits on body size imposed by the mechanics of flight, echolocation, and prey capture, may have resulted in reduction and simplification of the visual system as echolocation became increasingly important. The theory presented above is consistent with (and is indeed based on) bat monophyly and the phylogeny proposed in the current study. When considered in a phylogenetic context, the Eocene fossil bats Icaronycteris, Archaeonycteris,Hassianycteris, and Palaeochiropteryx are not informative about the origin of flight (which preceded diversification of the entire microchiropteran lineage), but they do provide some information concerning early steps in the acquisition of sophisticated echolocation. As noted above, some of the earliest morphological transformations in the lineage leading to extant Microchiroptera (those that took place before the divergence of Icaronycteris) included changes in features associated with the production of echolocation calls (enlargement of the cranial tip of the stylohyal), transmission of sounds through the middle ear (enlargement of the orbicular apophysis), and some fine-tuning of the inner ear (initial enlargement of the basal turn of the cochlea). Loosening of the attachment of the periotic to the basicranium evolved at this level or just subsequently (prior to the divergence of Archaeonycteris). It is interesting to note that all of these modifications occurred before achievement of what we described as ‘‘great enlargement of the cochlea’’ (i.e., to the size range of most extant microchiropterans), which evolved somewhat later, just prior to the divergence of Hassianycteris (fig. 40). Postcranial modifications that may have increased efficiency of the ventilation system (e.g., increased breadth of posterior laminae on the ribs) evolved in the microchiropteran lineage sometime between the divergence of Icaronycteris and the divergence of Palaeochiropterx. Other derived features potentially related to the echolocation system also evolved at some point in the early microchiropteran lineage. These include a phanerocochlear cochlea (probably associated in some fashion with cochlear expansion) and a deep, constricted fossa for m. stapedius (perhaps related to the system for avoidance of self-deafening). The former was minimally present before the divergence of Hassianycteris, the latter before the divergence of Palaeochiropteryx. As noted above, missing data for cochlear ossification in Icaronycteris and Archaeonycteris and for stapedial fossa form in Archaronycteris and Hassianycteris make it impossible to exactly place these transformations in the phylogenetic tree. Even given uncertainty about the relative timing of some transformations, our observations confirm the hypothesis that sophisticated echolocation evolved in a stepwise fashion in the early microchiropteran lineage. However, evolutionary changes in this system certainly did not cease with the origin of the microchiropteran crown group. Considerable cochlear size variation exists among living clades of Microchiroptera, with most extant microchiropterans (and all high-dutycycle echolocators) having a cochlea even larger than those seen in Palaeochiropteryx and Hassianycteris (Habserstezer and Storch, 1992). Reduction in cochlear size (back to the ‘‘moderately enlarged’’ condition) has apparently occurred in some lineages—in Mystacinidae, within Phyllostomidae, and within Megadermatidae (Fig. 40). Significant 1998 119SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS changes (some associated with cochlear enlargment) have also occurred in some lineages with respect to (1) length, width, and thickness of the basilar membrane, (2) length and height of the spiral ligament, (3) total number of cochlear neurons, (4) size and distributions of hair cell populations, (5) size and pattern of fluid spaces in the cochlea, and (6) form of the frequency map and ‘‘acoustic fovea’’ (Pye, 1966a, 1966b, 1967; Henson, 1970; Bruns, 1979; Bruns et al., 1981, 1983– 1984; Burda and Ulehlova, 1983; Ru¨bsamen et al., 1988; Neuweiler, 1990). Most or all of these changes may reflect modifications of the auditory system associated with different echolocation and foraging strategies (Bruns et al., 1981, 1983–1984; Ru¨bsamen et al., 1988; Neuweiler, 1990). FORAGING ECOLOGY OF EOCENE BATS One aspect of morphology and ecology not mentioned in the preceding discussion is the relationship between wing shape and foraging ecology. There is an extensive literature on this topic (e.g., Revilliod, 1916; Betz, 1958; Vaughan, 1959, 1966; Struthsaker, 1961; Hartman, 1963; Farney and Fleharty, 1969; Fenton, 1972; Findley et al., 1972; Kopka, 1973; Lawlor, 1973; Pirlot, 1977; Smith and Starrett, 1979; Norberg, 1981, 1986a, 1986b, 1987, 1994; Findley and Wilson, 1982; Findley and Black, 1983; Aldridge, 1986; Baagøe, 1987), with many recent contributions integrating data on echolocation call structure as well (e.g., Simmons et al., 1979; Neuweiler, 1984, 1989, 1990; Habersetzer, 1986; Aldridge and Rautenbach, 1987; Norberg and Rayner, 1987; Neuweiler and Fenton, 1988; Norberg, 1989, 1990, 1994; Fenton, 1990; Arita and Fenton, 1997). One outcome of this research has been identification of a series of features of wing design that affect flight performance and are correlated with foraging strategies. Among the most important measures of these are wing loading, aspect ratio, and a variety of wing tip indices designed to measure the size and shape of the dactylopatagium (e.g., Findley et al., 1972; Norberg and Rayner, 1987). Wing loading ( 5 body weight/wing area) provides a measure of the relative size of the wings and of the minimum weight per unit area that the wings must support during flight (Findley et al., 1972; Norberg and Rayner, 1987; Arita and Fenton, 1997). Increased wing loading requires greater flight speeds to generate enough lift to remain airborne; flight speed is proportional to the square root of wing loading (Findley et al., 1972; Norberg and Rayner, 1987). High wing loading is typical of fast fliers; slow-flying, more maneuverable animals usually have lower wing loading, with either lower body weight and/or increased wing area (Findley et al., 1972; Norberg and Rayner, 1987). Aspect ratio ( 5 wing span 2 /wing area; Norberg and Rayner, 1987) describes the overall shape of the wings by quantifying their length relative to their chord (Findley et al., 1972; Norberg and Rayner, 1987; Arita and Fenton, 1997). For any given body weight and wing loading, high aspect ratio wings are subject to less drag and thus facilitate greater flight speeds than do lower aspect ratio wings (Findley et al., 1972). However, high aspect ratio wings also generate less lift than do low aspect ratio wings, and thus bats with high aspect ratio wings may require greater wingbeat frequencies and airspeeds to remain airborne (Findley et al., 1972). Some bats with very high aspect ratio wings (e.g., Molossus,Eumops) cannot generate enough lift to become airborne from a flat surface (personal obs.), and thus usually drop from their elevated roosts to build up enough speed and lift to initiate flight (Vaughan, 1959). Low aspect ratio wings generate considerable drag at higher speeds, but maximize lift at low speeds (Findley et al., 1972). A variety of different measures of the relative size and shape of the wing tip have been proposed, including the tip index (Findley et al., 1972), alpha angle (Smith and Starrett, 1979), tip length ratio (Norberg and Rayner, 1987), tip area ratio (Norberg and Rayner, 1987), and tip shape index (Norberg and Rayner, 1987). All of these quantify various aspects of the relative size and shape of the dactylopatagium (that portion of the wing distal to digit V), which provides much of the propulsion generated by the wing during flight (Findley et al., 1972). The most comprehensive study of wing 120 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY design and foraging strategies in bats was that of Norberg and Rayner (1987), who summarized morphometric and behavioral data for more than 250 bat species. In addition to the ratios mentioned above, they also considered total body weight (mass), wingspan, and wing area. They interpreted these data in the context of mechanical and aerodynamic models for different modes of flight. Norberg and Rayner (1987: 337) summarized their results as follows: Some adaptive trends in bat wing morphology are clear from this analysis. Insectivores hunt in a range of different ways, which are reflected in their morphology. Bats hawking high-flying insects have small, pointed wings which give good agility, high flight speeds and low cost of transport. Bats hunting for insects among vegetation, and perhaps gleaning, have very short and rounded wingtips, and often relatively short, broad wings, giving good maneuverability at low flight speeds. Many insectivorous species forage by ‘flycatching’ (perching while seeking prey) and have somewhat similar morphology to gleaners. Insectivorous species foraging in more open habitats usually have slightly longer wings, and hence lower cost of transport. Piscivores forage over open stretches of water, and have very long wings giving low flight power and cost of transport, and usually long, rounded tips for control and stability in flight. Carnivores must carry heavy loads, and thus have relatively large wing areas; their foraging strategies consist of perching, hunting and gleaning, and wing structure is similar to that of insectivorous species with similar behavior. Perching and hovering nectarivores both have a relatively small wing area; this surprising result may result from environmental pressure for short wingspan or from the advantage of high speed during commuting flight; the large wingtips of these bats are valuable for lift generation in slow flight. Habersetzer and Storch (1987, 1989) and Habersetzer et al. (1994) capitalized on the observed relationships between wing form and flight behavior in extant bats to reconstruct the possible habits of Icaronycteris, Archaeonycteris,Hassianycteris, and Palaeochiropteryx. Habersetzer and Storch (1987) estimated wing loading, aspect ratio, and tip index for a series of fossil specimens, and compared these values with similar data from an extant bat fauna from tropical India (i.e., Habersetzer, 1986). They concluded that Icaronycteris and Archaeonycteris have a relatively high wing loading and low aspect ratio (Habersetzer and Storch, 1987, 1989; Habersetzer et al., 1994). Based on comparisons with extant forms, they found that Icaronycteris and Archaeonycteris have wings that resemble those of large rhinolophoids or the mean of vespertilionids (Habersetzer and Storch, 1987, 1989). Concerning the combination of low aspect ratio and high wing loading, Habersetzer and Storch (1989: 216) noted that ‘‘this parameter combination is most likely for unspecialized flight characteristics as can be found within the majority of the vespertilionids (e.g., Myotis myotis).’’ In contrast, Palaeochiropteryx was found to have a somewhat more specialized wing morphology (Habersetzer and Storch, 1987, 1989; Habersetzer et al., 1994). Palaeochiropteryx tupaiodon was characterized as having very low aspect ratio and very low wing loading, while both values were somewhat higher in P.spegeli (Habersetzer and Storch, 1987, 1989). On the basis of these values, both taxa were found to be very similar to extant rhinolophine and hipposiderine species (Habersetzer and Storch, 1987; Habersetzer et al., 1994). This similarity led Habersetzer and Storch (1987, 1989) and Habersetzer et al. (1994) to conclude that Palaeochiropteryx was characterized by slow, highly maneuverable flight close to the ground. In the case of P.tupaiodon, aerial capabilities may have included flight close to and even within foliage with potentially long-lasting hovering phases, as is seen today in Hipposideros bicolor, a species with a comparable combination of aspect ratio and wing loading values (Habersetzer and Storch, 1987, 1989; Habersetzer et al., 1994). Wing parameters of Palaeochiropteryx spiegali more closely resemble those of Hipposideros speoris, a form that hunts near obstacles but always stays in open airspace, never flying amongst foliage (Habersetzer and Storch, 1987, 1989; Habersetzer et al., 1994). Habersetzer and Storch (1987) found that both species of Hassianycteris are characterized by a high aspect ratio and high wing loading. On the basis of comparisons with Habersetzer’s (1986) data from Indian bats, Habersetzer and Storch (1987, 1989) and Habersetzer et al. (1994) observed that Hassianycteris species are even more specialized than high-flying tropical vespertilionids and rhinopomatids, instead showing a greater resemblance to extant high-flying molossids and emballonurids. As a result, they con- 1998 121SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS cluded that Hassianycteris messelensis and H.magna ‘‘can be considered to be distinctly adapted to flight in free spaces’’ (Habersetzer et al., 1994: 238). Habersetzer and Storch (1987, 1989) and Habersetzer et al. (1994) noted that all four Eocene genera are characterized by a short wing tip and a small dactylopatagium that accounts for only about 37% of the entire wing area. Among extant bats, they noted that this morphology is seen in rhinolophoids. Habersetzer et al. (1994) concluded that the Messel bats had succeeded in occupying a diverse set of ecological niches by evolving variations of the rhinolophoid wing type, and suggested that this wing type (i.e, low aspect ratio with short wing tip) represents the primitive wing form for bats. To further investigate the correlations described above, Habersetzer and Storch (1989) used multivariate methods developed by Norberg and Rayner (1987) to remove the effects of size (the first principal component), resulting in a plot of normalized wing loading (second principal component) versus normalized aspect ratio (third principal component). Although they used Norberg and Rayner’s (1987) method, Habersetzer and Storch (1989) did not use their data, but instead compared the fossil forms to Habersetzer’s (1986) reference fauna (with known flight and foraging habits) from tropical India. Habersetzer et al. (1989) interpreted the results of this multivariate analysis as supporting their previous conclusions about the flight behavior of the fossil bats (see above). However, Icaronycteris,Archaeonycteris trigonodon, and Palaeochiropteryx tupaiodon appeared to have very similar wing parameters based on the size-normalized analysis (at least in our judgment). This was not explicitly discussed by Habersetzer et al. (1989); however, they cautioned that their comparisons might be biased by the fact that they calculated body weight of the fossil forms using a measure of wingspan, which introduced circularity into the analysis and might have resulted in the observed clustering. Norberg (1989) reanalyzed wing morphology of the Eocene fossil forms in the context of the much larger data set collected by Norberg and Rayner (1987). Norberg (1989) suggested that the estimates of body mass calculated by Habersetzer and Storch (1987) may have been too high for several reasons (e.g., use of wingspan length regression equations, straight rather than curved wingtips in the fossil reconstructions). Using revised wing reconstructions (fig. 42) and revised estimates of body weight based on regression equations for radius length against total mass in recent bats, Norberg (1989) recalculated wing loading and aspect ratio for each of the fossil species, and additionally calculated wingtip length ratio, wingtip area ratio, and wingtip shape index for each taxon. Comparisons with recent bats (figs. 43, 44; data from Norberg and Rayner, 1987) led Norberg (1989: 204–205) to the following conclusions regarding Icaronycteris,Archaeonycteris,Hassianycteris, and Palaeochiropteryx: Based on my reconstructions, the ancient bats examined here . . . had low aspect ratio and high or average wing loading . . . (that is, high or average wing loadings in relation to body size), as compared with recent bats. This indicates that they had expensive and average to fast flight. Their wingtips were extremely short and with fairly large area, which are adaptations for maneuverable flight. Wingtip length ratios T 1 (handwing length/armwing length) are all between 0.93 and 1.09 ....T 1 , 1 is rather unusual among modern bats. The wing shape of the fossil species are similar to several recent pteropodids and phyllostomids and some rhinolophids (Rhinolophus hipposideros and R. ferrumequinum) and vespertilionids (Barbastella barbastellus,Lasiurus borealis,Rhogeesa tumida and some Myotis,Eptesicus and Pipistrellus species). Palaeochiropteryx and Hassianycteris were also similar to Rhinopoma hardwickei ..., both in wing shape and with their extremely short wingtips. Rhinopomatids are found mainly in deserts and steppes. They have been noted to forage for insects in open country as well as in open spaces around tree canopies, and their flight has been described as swift, fast and undulating (alternating flutters and glides)(see Norberg & Rayner 1987 for references) . . . Hassianycteris messelensis and H.magna had higher wing loadings and slightly higher aspect ratio than the other fossil species, and were thus faster fliers. They had though much lower aspect ratio than molossids and emballonurids. They may have been foraging in rather open spaces, like rhinopomatids, or along vegetation like noctules and lasiurines. The other ancient bats included here have lower aspect ratio and lower wing loading. Because their short wings and low aspect ratio these ancient bats probably foraged or lived among vegetation (which was also suggested by Habersetzer & Storch 1987) and may have been perch hunters. Perch hunting—making short flights out to 122 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY Fig. 42. Reconstructions of wing shapes of fossil bats; redrawn from Norberg (1989: fig. 3). Norberg (1989) modified reconstructions published by Habsersetzer and Storch (1987) by adding curved wingtips such as those seen in extant bats, changing the angles between the phalanges of the fourth digit in Archaeopteropus, and changing the angles between the third, fourth, and fifth digits in most species. capture prey detected from a fixed perch—is widely regarded as a behavior that reduces the energy required for successful foraging (Norberg and Rayner, 1987; Fenten, 1990; Fenton et al., 1990; Norberg, 1994). Extant bats that forage near or within vegetation may use perches because their short, low aspect ratio wings make flight relatively expensive (Norberg and Rayner, 1987; Norberg, 1994). Because flights to capture prey are infrequent and of short duration, less energy is apparently spent than would be required by sustained hawking (Norberg and Rayner, 1987; Fenton, 1990; Fenton et al., 1990; Norberg, 1994). Another possible advantage of perch hunting is that it may broaden the available prey spectrum by facilitating the capture and handling of larger 1998 129SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS sects while in slow flight. Palaeochiropteryx apparently specialized in preying on smallscaled insects (i.e., small moths and caddis flies), and its echolocation calls were most likely broadband calls of moderate to high frequency (30–90 kHz). There is no evidence that any of the Messl bats used high-dutycycle CF echolocation. Wing morphology of Icaronycteris and Archaeonycteris suggests that these taxa probably foraged close to the ground and close to vegetation, as did Palaeochiropteryx (Habersetzer and Storch, 1987, 1989, 1992; Norberg, 1989; and Habersetzer et al., 1992, 1994). However, presence of only moderate enlargement of the cochlea—together with absence of a calcar—suggests that these forms may have been perch hunters that specialized in gleaning their prey from surfaces rather than catching it on the wing. Archaeonycteris may have been a beetle specialist; we have no record of the preferred prey of Icaronycteris. Echolocation calls in both taxa were probably short ( # 2 msec) broadband FM signals of moderate to high frequency (30–90 kHz) or short, multiharmonic calls. In either case, these echolocation calls may have been of low intensity, and may have been ‘‘turned off’’ at times to facilitate location of prey by passive means (e.g., listening for prey-generated sounds or looking for prey movements). It is unlikely that Icaronycteris and Archaeonycteris used echolocation for detection, tracking, or evaluation of prey. Echolocation was problably used only for orientation and obstacle detection, while prey detection and tracking were accomplished by passive means. EVOLUTION OF FORAGING STRATEGIES: A PHYLOGENETIC PERSPECTIVE The phylogeny generated in our study (fig. 36) provides a framework for interpreting the morphological data and behavioral inferences presented above. Given the topology of this tree, we suggest that the earliest members of the microchiropteran lineage (forms currently unknown from fossils) probably used vision for orientation and obstacle detection in their arboreal/aerial environment, and probably foraged by gleaning insects and perhaps some fruits and other edible items from foliage, bark, and perhaps the ground. Location of potential food items was probably accomplished by a combination of vision and listening for prey-generated sounds. Like Archaeopteropus,Icaronycteris,Archaeonycteris, and Palaeochiropteryx, these bats probably had low aspect ratio wings, moderate wing loading, and relatively large wing tips, all of which suggest that they habitually flew close to the ground and near vegetation. Low-duty-cycle echolocation evolved subsequently, probably from communication calls that incidentally produced informative echoes. As we suggested earlier, this system was probably simple at first, permitting only orientation and obstacle avoidance but not detection, tracking, or evaluation of airborne prey. Basicranial modifications presumed to be associated with increased efficacy of echolocation began prior to the divergence of Icaronycteris from the microchiropteran lineage. However, wing morphology, a moderately enlarged cochlea, and absence of a calcar in this taxon suggest to us that Icaronycteris was a perch-hunting gleaner rather than a predator on aerial insects. This foraging strategy would have had the advantage of being relatively energy-efficient while at the same time requiring only moderate auditory data-processing capabilities to successfully sort the information from returning echoes. Echolocation calls were most likely short ( # 2 msec) broadband FM signals or multiharmonic calls, probably of relatively low intensity. These calls may have been ‘‘turned off’’ at times to facilitate passive prey localization. Indeed, prey detection, tracking, and evaluation were probably not done with echolocation, but rather by vision or listening for prey-generated sounds. Passive acoustic cues may have been particularly important if these bats were strictly nocturnal. In essence, the only major change in foraging method at this level would have been the addition of echolocation as a tool for orientation and obstacle detection. The basic foraging strategy—gleaning from a perch—would have been the same as seen in the nonecholocating chiropteran ancestors of Icaronycteris. The derived morphological transformations that diagnose the node linking Archaeo- 130 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY nycteris with Hassianycteris,Palaeochiropteryx, and the microchiropteran crown group (e.g., reduction of the number of roots on P3, presence of a ventral accessory process on C4, absence of an ossified third phalanx on wing digit III) are relatively small changes that do not indicate any major shifts in foraging ecology. This pattern suggests to us that Archaeonycteris retained much the same foraging strategy as Icaronycteris.Archaeonycteris and Icaronycteris share a similar wing morphology, moderately enlarged cochlea, and lack of a calcar, features we interpret as indicating that these bats foraged by gleaning prey that were detected from a perch using passive means rather than echolocation. The next node as one moves up the tree— that which links Hassianycteris with Palaeochiropteryx and the microchiropteran crown group—is associated with a much more extensive suite of morphological changes (see discussion above under ‘‘Character Transformations at Basal Nodes’’). Most notable among these are: (1) a greatly enlarged cochlea, (2) a dorsal articular facet on the scapula (although this may have evolved earlier), (3) a laterally compressed ventral process on the manubrium of the sternum, (4) increased development of rib laminae (this may have evolved somewhat earlier), (5) a threadlike fibula, and (6) presence of a calcar. Taken together, these features suggest that a major shift in foraging strategy occurred in the microchiropteran lineage just prior to the divergence of Hassianycteris: the evolution of aerial hawking. This behavior involves using echolocation to detect, track, and assess prey, and the use of the uropatagium (supported and controlled by a calcar) to capture prey on the wing. Changes in the postcranial skeleton, particularly the pectoral girdle, suggest that some ‘‘fine tuning’’ of the flight mechanism may have accompanied the behavioral change to aerial hawking. This seems reasonable given the demands of aerial foraging behavior, particularly if these bats occasionally foraged in group situations (e.g., when caddis flies were hatching along the shore of Lake Messel) where complex capture and evasive maneuvers may have been required. Interestingly, the morphological and behavioral changes associated with the evolution of aerial hawking do not seem to have been correlated with any appreciable modifications in wing shape or wing loading (based on the results of Norberg,1989; see figs. 43, 44). Similarities in aspect ratio, wing tip indices, and wing loading in Palaeochiropteryx,Archaeonycteris,Icaronycteris, and Archaeopteropus lead us to conclude that the body size and wing form of Palaeochiropteryx is plesiomorphic, and that this morphology was present in the most recent common ancestor of Palaeochiropteryx and Hassianycteris. Accordingly, the body size and wing morphology seen in Hassianycteris (discussed below) are autapomorphic features. Like Icaronycteris and Archaeonycteris, Palaeochiropteryx had low aspect ratio wings, moderate wing loading, and relatively large wing tips, all of which indicate that these bats habitually flew close to the ground and near vegetation. However, the greatly enlarged cochlea and well developed calcar suggest that Palaeochiropteryx was fully capable of aerial hawking. This hypothesis is consistent with the analyses of stomach contents and taphonomy of Palaeochiropteryx at Lake Messel. Echolocation signals used by Palaeochiropteryx were probably shortto moderate-length broadband FM calls. As pointed out by Fenton (1990) and Fenton et al. (1990), many extant bats that hunt from perches also hawk insects in continuous flight, with allocation of time between these two foraging strategies reflecting prey availability. Given that gleaning from a perch probably represents the primitive foraging strategy for the microchiropteran lineage, we suspect that Palaeochiropterx may have used a combination of perch hunting (including flycatching) and slow aerial hawking to capture its prey, much like modern nycterids. This would be consistent with cochlear size in Palaeochiropterx (as estimated by Habersetzer and Storch, 1992; fig. 29), which places this genus just at the lower end of the range of variation seen among forms that are continuous aerial hawkers (e.g., most vespertilionids) and at the upper end of the range of variation in extant perch-hunting forms that use low-duty-cycle echolocation (e.g., nycterids and megadermatids). The transition from gleaning stationary 1998 131SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS prey detected by passive means (vision or listening for prey-generated sounds) to aerial hawking using echolocation to detect and track prey must have been a complex process that involved intermediate steps. It seems unlikely that echolocation was used for detecting stationary prey during this transition, because detection of hard targets resting on hard or irregular backgrounds is one of the more difficult tasks faced by echolocators (Simmons et al., 1980; Fenton, 1990, 1994a, 1995). Instead, we hypothesize that bats that regularly foraged by gleaning stationary prey (detected by passive means) increasingly came to hunt moving insects, perhaps leaping after prey that had been startled by movements or attacks by the bat. At this point, echolocation (which they were already using for orientation and obstacle avoidance) could provide more information about prey location and movement than could be obtained by passive means. To fully exploit this form of data collection, the animals would have had to increase signal strength in order to maximize range and provide the necessary time to track and evaluate targets. However, environmental clutter (which produces many distracting echoes) would have presented a major impediment to the evolution of effective flycatching behavior. It therefore seems likely that flycatching using echolocation to detect and track prey probably evolved in bats that frequented forest gaps and the edges of forests along lakes and rivers, places where vegetation (with potential perches) lies adjacent to relatively clutter-free open spaces. Once flycatching from perches was well established, it is easy to imagine a progressive transition to spending more time on the wing, ultimately leading to the evolution of taxa that relied exclusively on foraging by continuous aerial hawking. It is not clear when bats began to use the uropatagium for prey capture, but evolution of the calcar apparently preceded or was coincident with the evolution of aerial hawking. Our hypothesis that the evolutionary transition from gleaning (using passive prey detection) to aerial hawking (using echolocation) took place in habitats associated with forest gaps or forest edges along bodies of water cannot be tested given the sparse fossil record of bats. However, it is interesting to note that all four of the fossil bats considered here (Icaronycteris,Archaeonycteris,Hassianycteris, and Palaeochiropteryx) have been collected from lake deposits that are presumed to have been surround by subtropical or tropical forests at the time of deposition (MacGinitie, 1969; Grande, 1980; Schaarschmidt, 1992). The excellent preservation of many of the bat fossils found at Messel and Fossil Basin indicates that these animals were not transported far after death, suggesting that they may have foraged around the edges and over the lakes in which they were ultimately preserved. Species of Hassianycteris are larger in many dimensions than most of the other Eocene bats, and analyses of wing morphology by Norberg (1989) indicated that Hassianycteris was characterized by significantly higher wing loading than were Palaeochiropteryx,Archaeonycteris,orIcaronycteris. Indeed, Hassianycteris apparently had a higher wing loading than most extant bats (fig. 43). These observations, taken together with a reasonably high aspect ratio, a greatly enlarged cochlea, and presence of a calcar, suggest that Hassianycteris foraged by fast aerial hawking, most likely well above the ground in forest gaps or above the canopy. Hassianycteris thus represents another shift in foraging strategy, away from combined perch hunting and aerial hawking near the ground or vegetation to a fast-flying, continuous aerial hawking foraging strategy similar to that seen in extant rhinopomatids and some vespertilionids. As noted above, however, this shift apparently occurred after the lineage leading to Hassianycteris diverged from the lineage leading to the microchiropteran crown group. Continuing to move up the phylogenetic tree, we find that the clade comprising Palaeochiropteryx plus the microchiropteran crown group is diagnosed by only one unambiguous synapomorphy, the presence of a ventral accessory process on C5. This feature is not indicative of any change in foraging habits, but rather a continuation of neck modifications associated with roosting behavior (see discussion under character 77). Transformations that diagnose the microchiropteran crown group (e.g., modification of the premaxilla articulation, reduction in the 132 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY number of lower premolars) indicate changes in the masticatory apparatus, but no major behavioral shifts. In summary, we propose that foraging behavior in the microchiropteran lineage evolved in a series of steps: (1) gleaning food objects during short flights from a perch using vision for orientation and obstacle detection; prey detection by passive means, including vision and/or listening for prey-generated sounds (no known examples in fossil record); (2) gleaning stationary prey from a perch using echolocation and vision for orientation and obstacle detection; prey detection by passive means (Icaronycteris;Archaeonycteris); (3) perch hunting for both stationary and flying prey using echolocation and vision for orientation and obstacle detection; prey detection and tracking using echolocation for flying prey and passive means for stationary prey (no known example, although Icaronycteris and/or Archaeonycteris may have done this at times); (4) combined perch hunting and continuous aerial hawking using echolocation and vision for orientation and obstacle detection; prey detection and tracking using echolocation for flying prey and passive means for stationary prey; calcar-supported uropatagium used for prey capture (common ancestor of Hassianycteris and Palaeochiropteryx; retained in Palaeochiropteryx); and (5) exclusive reliance on continuous aerial hawking using echolocation and vision for orientation and obstacle detection; prey detection and tracking using echolocation (Hassianycteris; common ancestor of microchiropteran crown group). Given the topology of the tree we derived earlier (fig. 36), it seems most likely that the latter foraging strategy—reliance on continuous aerial hawking—was primitive for the microchiropteran crown group. This is consistent with optimization of foraging strategies both within the microchiropteran crown group and among the fossil stem group forms. The conclusion that aerial hawking is the primitive foraging strategy for the microchiropteran crown group suggests that gleaning, passive prey detection, and perch hunting among extant taxa represent secondarily derived specializations rather than retentions of primitive habits. If so, topology of our tree (fig. 36) indicates that passive prey detection evolved independently at least five or six times—in phyllostomids, mystacinids, antrozoids, vespertilionids, and either independently in nycterids and megadermatids or in the common ancestor of Rhinolophoidae. Similarly, gleaning must have evolved independently at least four times within Microchiroptera—in phyllostomids, antrozoids, vespertilionids, and rhinolophoids. Perch hunting apparently evolved at least three times—in phyllostomids, vespertilionids, and rhinolophoids. Although passive prey detection, gleaning, and perch hunting are linked in some taxa (e.g., most megadermatids and nycterids), they are decoupled in other forms. For example, Mystacina apparently uses passive prey detection but does not glean or hunt from perches; instead, it approaches its prey ‘‘on foot’’ (B. Lloyd, personal commun.). Rhinolophids sometimes hunt from perches and glean, but apparently do not use passive cues to detect their prey (Norberg and Rayner, 1987). Lavia frons hunts from perches but apparently does not glean or use passive prey detection (Vaughan and Vaughan, 1986). These observations suggest that passive prey detection, gleaning, and perch hunting evolved in different ways in different microchiropteran lineages. The relative timing of the evolutionary and behavioral changes in the lineage leading to extant Microchiroptera can be estimated from the fossil record. Taxa representing stages 2, 4, and 5 as defined above were apparently present simultaneously at Messel, suggesting that evolutionary transformations in foraging strategies may have occurred very rapidly in early members of the microchiropteran lineage. This was probably facilitated by the mechanical coupling of ventilation and flight, which meant that the energy costs of echolocation to flying bats were low, particularly in comparison to the energetic benefits of aerial hawking for insects in the absence of any competitors. The evolution of continuous aerial hawking may have been the ‘‘key innovation’’ (sensu Liem, 1973) responsible for the burst of diversification in microchiropteran bats that occurred in the Eocene. Fossils referable to six major extant lineages are known from Middle–Late Eocene deposits (table 1): (1) 1998 133SIMMONS AND GEISLER: RELATIONSHIPS OF EOCENE BATS Emballonuridae, (2) Megadermatidae, (3) Hipposiderinae, (4) Rhinolophinae, (5) Nataloidea, and (6) Molossinae. Reconstruction of ghost lineages following the methods of Norell (1992) leads to the conclusion that many more extant lineages were minimally present by the end of the Eocene, including (7) Rhinopomatoidea, (8) Nycteridae, (9) Noctilionoidea, (10) Mystacinidae, (11) Antrozoidae, (12) Tomopeatinae, and (13) Vespertilionidae. All of these must have diverged during the Eocene given the Eocene ages of their sister-taxa. It thus appears that Icaronycteris and the Messel bats provide an unprecedented view of the steps leading up to a major adaptive radiation of mammals. CLASSIFICATION OF EOCENE BATS The phylogenetic results of this study indicate that many groupings of Eocene taxa previously recognized in formal classifications (e.g., Eochiroptera, Palaeochiropterygoidea, Archaeonycterididae [including Icaronycteris]) are not monophyletic. Beginning at the lowest taxonomic level, it seems most appropriate to place each of the four genera considered in this study in its own monophyletic family 13 :Icaronycteris in Icaronycteri13 There is considerable disagreement in the literature concerning the formation of family-group names based on generic names ending in -nycteris. The family name in common usage for extant Nycteris is Nycteridae (Koopman and Jones, 1970; Hill and Smith, 1984; Koopman, 1984, 1993, 1994). Revilliod (1917b) used a similiar formation when he named Archaeonycteridae, as did Jepsen (1966) when he named Icaronycteridae. However, Russel and Sige´ (1970) argued that the proper formation of family-group names from -nycteris required the spelling -nycterididae. They thus changed the spelling of Archaeonycteridae to Archaeonycterididae. Some authors have followed this usage (e.g., Habersetzer and Storch, 1987, 1989; Habsersetzer et al., 1992, 1994), but many have continued to use Archaeonycteridae (e.g., Hill and Smith, 1984; Hand et al., 1994; McKenna and Bell, 1997). Habsersetzer and Storch (1987) followed Russel and Sige´’s (1970) usage when they named Hassianycterididae, and subsequent authors have used this spelling (e.g., Habersetzer et al., 1992; McKenna and Bell, 1997). Despite Russell and Sige´’s (1970) argument about the nature of the greek root of -nycteris, we think that it is counterproductive to spell some family-group names in one fashion and others differently. Because Nycteridae is widely accepted as the spelling of the family-group name based on Nycteris, we argue that all family-group names based on genera ending in -nycteris should be spelled in the same fashion. Accordingly, we recognize the Icaronycteridae, Archaeonycteridae, and Hassianycteridae as the most appropriate spellings for familygroup names based on Icaronycteris,Archaeonycteris, and Hassianycteris. dae Jepsen, 1966; Archaeonycteris in Archaeonycteridae Revilliod, 1917b; Hassianycteris in Hassiancyteridae Habersetzer and Storch, 1987; and Palaeochiropteryx in Palaeochiropterygidae Revilliod, 1917b. This arrangement preserves monophyly of higher taxonomic groups and serves to highlight the morphological and presumed behavioral differences among genera. As we view them, Icaronycteridae, Archaeonycteridae, and Hassiancyteridae each currently contain only the nominate genus. The situation is slightly more complex in the case of Palaeochiropterygidae. Two additional genera, Cecilionycteris and Matthesia, have been referred to Palaeochiropterygidae by previous authors based on dental features (e.g., Sige´ and Russell, 1980; Hill and Smith, 1984; Beard et al., 1992). In the absence of additional data, we provisionally accept this assessment, although we note that discovery of more complete material may ultimately demonstrate that Palaeochiropterygidae as thus defined (including Palaeochiropteryx, Cecilionycteris, and Matthesia) is a paraphyletic assemblage. Ageina, which is known only from dental fragments, has been referred variously to ‘‘family uncertain’’ (Russell et al., 1973), Palaeochiropterygoidea (Smith, 1977; Smith and Storch, 1981), Eochiroptera (Sige´ and Legendre, 1983), Archaeonycteridae (Hill and Smith, 1984), and possibly Natalidae sensu Van Valen (Beard et al., 1992). We consider the assessment of Beard et al. (1992), which is based on the most recent evidence (including comparisons with the new taxon Honrovits), to represent the best current working hypothesis. Accordingly, we follow suggestions made by Beard et al. 134 NO. 235BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY (1992) and refer both Ageina and Honrovits to Nataloidea incertae sedis. Australonycteris, which is also poorly known, was placed in Microchiroptera, family incertae sedis by Hand et al. (1994). They noted that this form has a dentition similar to that of Archaeonycteris (at least in terms of the absence of derived traits), although the petrosal referred to Australonycteris appears relatively more derived and is somewhat vespertilionoid-like (Hand et al., 1994). These and other observations suggest to us that Australonycteris probably fits in the tree somewhere between Archaeonycteris and the microchiropteran crown group. Given this hypothesis, any decision about how to classify Australonycteris requires consideration of broader issues surrounding the classification of stem-group forms. The principal nomenclatural problem faced in this study concerns higher-level classification of the Eocene groups that fall outside the microchiropteran crown group— in other words, where to put Icaronycteridae, Archaeonycteridae, Hassiancyteridae, and Palaeochiropterygidae. Debates have raged in the systematic literature for decades concerning the relative pros and cons of different approaches to defining and naming taxonomic groups (e.g., Ghiselin, 1966, 1984; Nelson, 1972, 1974; Bock, 1974; Duncan and Estabrook, 1976; Estabrook, 1978, 1986; Jefferies, 1979; Wiley, 1979, 1981; Duncan, 1980; Phillips, 1984; Rowe, 1987, 1988; Gauthier et al., 1988; Heywood, 1988; de Quieroz and Gauthier, 1990, 1992, 1994; Minelli, 1991; Lucas, 1992; Meier and Richter, 1992; Lucas and Luo, 1993; Wyss and Flynn, 1993; Bryant, 1994, 1996; de Queiroz, 1994; Smith, 1994; Wyss and Meng, 1996; McKenna and Bell, 1997). Four principal issues have been debated: (1) whether taxa should be recognized on the basis of shared characters (i.e., diagnoses) or defined phylogenetically (i.e., on the basis of their relationships); (2) whether all named groups above the species level must be monophyletic (some workers have argued that convex paraphyletic groups may be usefully employed in classifications); (3) the pros and cons of recognizing formal ranks (e.g., ‘‘order,’’ ‘‘family’’) in classifications; and (4) the relative merits of nodebased, crown-clade-restricted, stem-based, stem-modified node-based, and other types of phylogenetic definitions. We concur with most recent authors in concluding that phylogenetic definitions are essential and that all named taxa should be monophyletic (or at least potentially monophyletic). We consider issues related to rank to be of secondary importance compared with the definition of taxa (see discussion below). The central issue, in our opinion, is the method(s) used to define the limits of taxa, particularly those to which we may apply well-known taxonomic names (e.g., Microchiroptera). Given the phylogenetic relationships hypothesized here, three general options for defining Microchiroptera seem appropriate. Using a node-based definition that emphasizes the importance of the extant crown group (i.e., a crown-clade-restricted definition), we might define Microchiroptera as the clade stemming from the most recent common ancestor of Emballonuridae and Yangochiroptera. If defined this way, Yinochiroptera would be included within Microchiroptera regardless of whether Emballonuridae falls outside Yinochiroptera (as suggested by this study) or inside Yinochiroptera (as suggested by Koopman, 1985, 1994). A less explicit crown-clade definition might define Microchiroptera as the clade stemming from the most recent common ancestor of all extant bats that use sophisticated echolocation. When defined either way, Microchiroptera would be equivalent to the microchiropteran crown group as discussed earlier, and would therefore exclude Palaeochiropterygidae, Hassiancyteridae, Archaeonycteridae, and Icaronycteridae. The main advantages of these definitions are that they are congruent with general usage of the name Microchiroptera by most biologists, and that they can be expected to remain relatively stable (e.g., discovery of new fossils will not affect the limits of Microchiroptera; fossil species would either fall inside or outside this clade). The latter option (defining Microchiroptera as the clade stemming from the most recent common ancestor of all extant echolocating bats) depends on persistence of extant forms, and thus might be considered potentially unstable because major extinctions could affect group contents in the future (see discussion of this problem in Lucas [1992] and Lucas and Luo [Document text truncated for crawler view.]