Skip Navigation
  • UC Riverside
  • College of Natural and Agricultural Sciences

The Comparative Method and Why Phylogeny Matters Some talks from the 29th Annual North American Symposium on Bat Research
     held 27-30 October 1999 at the University of Wisconsin-Madison.

These 8 talks were in a session entitled "Bat Phylogeny and the Comparative Method"
     held Friday the 29th 2:00 - 4:30 P.M.

The session was co-organized by Theodore Garland, Jr. and Ariovaldo Pereira da Cruz-Neto.

The comparative method and why phylogeny matters

Theodore Garland, Jr.

University of Wisconsin-Madison

The last 15 years have witnessed a revolution in the way species differences are studied: the "comparative method" has been revitalized by new analytical tools that use phylogenies and by increased phylogenetic information. Most typically, modern analyses obtain information about one or more phenotypic traits (e.g., wing area, metabolic rate, relative brain size, frequency of echolocation calls, social system, diet, home range area) for a series of species, and then "map" this information onto a phylogenetic tree that has been obtained from independent data (e.g., DNA sequences), analyzed with an appropriate tree-reconstruction algorithm. Usually, to avoid circularity, the traits of interest are not used to construct the phylogeny. Phylogenetically based statistical methods (review in Am. Zool. 39:374-388) attempt to account for the fact that related species tend to resemble each other, and hence that their phenotypes cannot be considered as independent and identically distributed data points for analyses. These techniques allow traditional topics in comparative and ecological physiology, functional morphology, and behavioral ecology to be addressed with greater rigor. Examples include studies of the form of allometric relationships, prediction of values for unmeasured species, correlated character evolution and coadaptation, and whether characteristics of organisms vary consistently in relation to behavior, ecology or environmental characteristics, which can constitute evidence of adaptation (results of natural selection). These methods can also address topics that are unapproachable without phylogenetic information, such as where and when a trait first evolved, its value in hypothetical ancestral species, and whether rates of evolution have differed among evolutionary lineages (clades). Although most powerful with complete phylogenetic information, the procedures can incorporate uncertainty about topology (polytomies: Syst. Biol. 48:547-558), branch lengths, and the way characters have evolved. Three general methods will be discussed: reconstruction of ancestral states (Evolution 51:1699-1711), phylogenetically independent contrasts (Am. Nat. 125:1-15), and Monte Carlo computer simulations to obtain phylogenetically correct null distributions (Syst. Biol. 42:265-292). Free software to implement these techniques is available from the author.

Phylogenies and where they come from: conflict, concurrence, and the comparative method

John A. W. Kirsch

The University of Wisconsin Zoological Museum

The types of characters and algorithms used to generate phylogenetic trees bear an intimate, and often unrecognized, relation to each other: not all characters are suited to the same analytical method, and vice versa. Parsimony, for example, has as its first assumption the independence of unit features, an assumption that may at least in principle be met by disparate anatomical characters but rarely by the sequence of bases in a single gene. On the other hand, analogues of the plausible models for DNA evolution - such as those accounting for the influence of transition:tranversion ratios, saturation, base-compositional bias, differential weighting of portions of a gene, etc. - are unknown for anatomy and enhance the power of computer methods appropriate for molecules. Yet, conservative assumptions that base changes are unordered and non-polarized effectively renders sequence analysis "phenetic" (in the sense of "based on overall similarity"). In this, the results resemble those of distance-generating methods (serology, DNA hybridization). As regards bat phylogeny, molecular trees confirm the monophyly of many family-level groups inferred from anatomy, and appear to "solve" several conundra such as the affinities of Mystacina; but they differ from tradition markedly at higher levels, especially concerning the time-honored subordinal dichotomy between megabats and microbats, uniformly recovering a sister-group relation between Pteropodidae and Rhinolophoidea. Because widely-applied combinability tests assess only length differences rather than branching orders among trees, and because few techniques are capable of combining data when taxa or characters differ among studies, it is premature to adopt the results of a global or "total-evidence" analysis until the algorithmic or other reasons for such conflicting results are understood. The more general lesson is that differing trees corresponding to various classes of data may have equal legitimacy. The dilemma this poses for comparative-methodologists is that whereas a single tree can be provided, it may not be the (most) correct one; several phylogenetic hypotheses need to be examined. However, the plausibility or otherwise of the reconstructions of character-evolution on multiple trees may contribute to a considered choice amongst phylogenies.

Bat phylogeny: refinements in the morphological perspective

William A. Schutt, Jr.1,2 and Nancy B. Simmons1

1 Department of Mammalogy, American Museum of Natural History, New York, NY
2 Natural Science Division, Southampton College of Long Island University, Southampton, NY

Although much recent systematic work has been based on molecular data, analysis of morphology continues to provide considerable insight into bat relationships. Recently published studies of higher-level bat phylogeny by Simmons and her colleagues have provided strong support for monophyly of many traditionally recognized groupings. However, the level of taxonomic sampling in these studies, which used families and subfamilies as the terminal taxa, resulted in significant polymorphism in the data set. Particularly problematic taxa include many of the large and diverse families (e.g., Pteropodidae, Phyllostomidae, Molossidae, and Vespertilionidae). Sampling at the family/subfamily level has also made combining morphological and molecular data sets difficult since in many cases, molecular data are sampled at the species level. To address these problems, we are assembling a new morphological data set designed to test higher-level bat relationships using multiple species-level exemplars. Taxa have been chosen so as to insure availability of morphological specimens (i.e., skins, skulls, and fluid-preserved specimens) and tissue samples from the same species. Our data set presently includes over 350 morphological characters sampled, as far as possible, in over 50 species representing all chiropteran families. Characters have been drawn from multiple anatomical systems including the musculoskeletal system, tongue, face and ears, pelage and patagia, reproductive tract, respiratory and digestive systems, and brain. Preliminary phylogenetic results are broadly congruent with previous studies, with some interesting exceptions. Future work on this project will include addition of data from more species, filling in data[ gaps in many hard-to-study anatomical systems, and investigation of new characters. In collaboration with the Van Den Bussche lab, we are also working to analyze our morphological data in the context of combined analysis with molecular sequence data from the same species.

Phylogenetic analysis of multiple independent data sets: to combine or not?

Ronald A. Van Den Bussche1, Steven R. Hoofer1, William A. Schutt, Jr.2,3, and Nancy B. Simmons3

1Department of Zoology, Oklahoma State University, Stillwater, OK 74078
2Natural Science Division, Southampton College of Long Island University, Southampton, NY
3Department of Mammalogy, American Museum of Natural History, New York, NY

With multiple independent data sets being generated for phylogenetic analyses, several questions are emerging regarding how they should be analyzed. For example, is it necessary for data sets to have similar numbers of characters to prevent "swamping out" the phylogenetic information in smaller data sets? Controversy also exists regarding how different data partitions should be analyzed. Three approaches (total evidence, separate analyses, and conditional combination) have been proposed for elucidating phylogenetic hypotheses for organisms based on multiple data partitions. Although several statistical methods exist for testing whether different data partitions can be combined as well as for testing congruence among phylogenetic hypotheses resulting from separate analyses, the performance of these tests has not been thoroughly examined nor have these tests been applied on a broad taxonomic scale. Therefore, little information is available regarding the prevalence of significant heterogeneity among data partitions. To elucidate higher-level phylogenetic relationships within Microchiroptera, for all taxa under study we are generating approximately 2.7 kb of DNA sequences from three adjacent mitochondrial genes and 350 morphological characters. Because our goal is to provide a well resolved, well-supported phylogenetic tree for Microchiroptera, it is necessary to compare the phylogenetic information present in each of these data sets. The specific purpose of this study was to evaluate the phylogenetic affinities of the New Zealand short-tailed bat (Mystacina tuberculata) based on molecular and morphological characters. Toward this end we (1) performed separate phylogenetic analyses and evaluated congruence among different data sets, (2) utilized three statistical tests of homogeneity for determining whether a "total evidence" approach could be utilized, and (3) examined the effect of data partitions of different size and evolutionary constraints on the resulting phylogenetic trees.

The phylogeny of echolocation

James M. Hutcheon, Theodore Garland, Jr., and John A.W. Kirsch

Department of Zoology and University of Wisconsin Zoological Museum, Madison

Ongoing controversies in bat systematics concern not only the interrelationships of chiropteran families, superfamilies, and sub-orders, but also the identity of the nearest likely outgroup of bats. Echolocation is frequently pointed to by all sides in this debate as relevant to their particular phylogenetic hypotheses. Under the standard or "flight first" hypothesis, where powered flight is considered the most general uniting feature of chiropterans, pharyngeal sonar is viewed as a synapomorphy of Microchiroptera, never having evolved in Megachiroptera; the "echolocation first" scenario, in contrast, mandates the loss of this capacity in megabats. Recent molecular work, however, challenges the categorical distinction between microbats and megabats, allying at least Rhinolophoidea with the Pteropodidae, and thus raises the possibility that echolocation evolved twice among chiropterans. But echolocation is not a unitary anatomical feature nor is it expressed in the same manner in all bats which possess it. The comparative method provides a means of assessing the evolution of traits by taking relationships - as expressed in a phylogenetic tree - into account, thus reducing spurious correlations caused simply by shared ancestry. Body mass, foraging strategy, and echolocation call-frequency are considered to be associated. Examining the hypothesis that foraging strategy in bats is related to body mass in a phylogenetic context provides an opportunity to account for what might, in fact, be a phylogenetic constraint. Likewise, although everyone "knows" that megachiropteran bats are larger than many microbats, does variance in body size represent a statistically significant difference, or do mean differences between suborders fall within the expected variance of an evolutionary process? Finally, reference is often made to the large number of extant bat species, with the explicit assumption that echolocation, flight, or both represent a key innovation(s) allowing for the diversification observed in bat species today. However, while the notion of echolocation as a key innovation has been marshalled as an argument to support all of the competing phylogenetic theories, it this hypothesis has not been the subject of a rigorous, statistical assessment.

Size matters: scaling and the evolution of flight and echolocation in bats

Hector T. Arita

Instituto de Ecologia, UNAM; Apdo. Postal 27-3 (Xangari); CP 58089 Morelia, Mich.; MEXICO

The maximum body mass for bats is three orders of magnitude lower than the weight of the largest terrestrial mammals and one order of magnitude below the theoretical limit for a flying vertebrate. I suggest that three types of constraints could be limiting the size of bats: (1) those associated with flight; (2) those associated with echolocation, especially the coupling of flight and echolocation in aerial-hawking species; (3) phylogenetic constraints. Flight alone might limit the size of megabats, but other constraints (associated with diet, roosting habits or viviparity) could be acting as well. At least in some aerial insectivorous microbats, wing flapping and echolocation sound production are coupled to save energy, so one echolocation call corresponds in time to one wing beat. In theory, pulse repetition rate (PRR) and wing-beat frequency (WBF) should scale with body mass with the same exponent to yield PRR/WBF ratios equal to one. However, the exponents vary among taxa, showing that some bats donít couple the two functions. I traced changes in body mass, PRR, and WBF along N. Simmonsí phylogeny of bats to infer the possible effects of diet, echolocation call design, and phylogenetic constraints on body size among different bat lineages. I found three generalized patterns: First, high-intensity echolocation and its coupling with flight constrain size in aerial-feeding insectivores. Second, in lineages with species using low-energy echolocation ("whispering" bats), some species have attained comparatively large sizes, notably among frugivores and carnivores. However, even these bats are much smaller than the largest Megachiroptera. Finally, lineages of bats using high-duty, constant-frequency echolocation calls show a secondary reduction in body size. I conclude that: (1) flight constrains size in all bats, (2) an additional constraint must be limiting the size of Megachiroptera, (3) size in Microchiroptera is constrained by echolocation, (4) aerial insectivores are particularly small, apparently because of energetic constraints, (5) size in frugivores and gleaners are limited to a lower degree by echolocation constraints, and a phylogenetic constraint might be acting, (6) a different mechanism must be limiting body size in bats with high-duty echolocation.

Bat life-histories: testing models of life-history evolution in mammals using a comparative phylogenetic approach

Kate E. Jones* and Ann MacLarnon

School of Life Sciences, Roehampton Institute London, London, SW15 3SN, UK

The evolution of bat life-histories (Mammalia: Chiroptera) was examined to investigate patterns of life-history covariation and to test Charnovís (1991) mammalian model of life-history evolution. Life-history data was collected from the literature for 308 bat species. Data were tranformed into phylogentically independent data using Felsensteinís (1985) method of independent comparisons, modified by Pagel (1992) and implemented by the computer program CAIC (Purvis & Rambaut 1995). As no single estimate of the phylogenetic relationships for all the taxa under investigation was available, we calculated a ëphylogenetic supertreeí (Sanderson et al. 1998) from 53 different systematic studies to cover the taxa required. Branch lengths in the phylogeny were set to the same arbitrary value in the CAIC program and the adequate standardisation of the contrasts was checked prior to analysis (Garland et al. 1992). With phylogenetically independent contrasts, several of the bat life-history variables investigated were correlated with body mass but others were independent of body mass. The allometric scaling of the life-history variables was not significantly different within the two suborders of bats. Covariation of bat life-history traits was also examined. Although differences were found in the pattern of bat life-history covariation compared to other mammals, fundamental similarities remain; for example, low mortality rates are correlated with late ages at maturity. This means that the ëfast-slowí continuum found to operate in other mammals is also found among bat species. The life-history covariation found within bats supports, or at least not strongly refute, Charnovís (1991) model. Charnovís predicted values for the allometries and interrelationships of juvenile period length (a) and adult mortality (M), and his predicted invariants, are supported for bat species. However, annual fecundity (b) is not significantly correlated with body size as Charnov predicts. A more recent life-history model developed by Kozlowski and Weiner (1997) may explain some of the patterns observed for bat life-histories but empirical analyses have yet to be performed.

*Current Address: Department of Biology, Imperial College at Silwood Park, Ascot, Berkshire, SL5 7PY, UK

The relationship between body mass, phylogeny, diet, and basal metabolic rate in phyllostomid bats:
a computer-simulation approach

A. P. Cruz-Neto and A. S. Abe

Departamento de Zoologia, IB, UNESP, CP 199, CEP 13506-900, Rio Claro, SP, Brasil

Controversy exists as to what factors affect the basal metabolic rate (BMR) in mammals, once the pervasive effects of body mass have been controlled. For example, various authors have suggested that BMR varies also in relation to diet or phylogeny. We examined these factors in the bat family Phyllostomidae. We first used conventional analysis of covariance (ANCOVA) to test the effects of diet on BMR, while controlling for effects of body mass. This conventional analysis assumes that all species evolved simultaneously from a single common ancestor (i.e., a star phylogeny with equal-length branches) and a Brownian motion model of character evolution. In this analysis, diet exerted a strong effect on mass-corrected BMR. We next used a phylogenetically informed analysis involving Monte Carlo simulations along a specified phylogenetic tree and under different models of character change. The PDSIMUL computer program (Syst. Biol., 1993, 42:265-292) was used to simulate the evolution of body mass and BMR thousands of times, and these data were then analyzed in the same way as for the one set of real data. The F statistic for the real data set was compared to the distribution of F statistics from the computer-simulated data. In this analysis, no effect of diet on BMR was observed, irrespective of the evolutionary model assumed. In both conventional and phylogenetically informed analyses, body mass exerted a strong effect on BMR: larger-bodied species had lower mass-specific metabolic rates. Owing to the fact that diet is perfectly confounded with phylogeny in the set of 19 species examined, even if diet did exert an effect on BMR in the phylogenetic analysis, it would not be possible to separate this effect from the possibility that these phyllostomid bats differ for some other reason besides diet. This points to the need for careful selection of study species if one hopes to elucidate the effects of ecological or behavioral factors on aspects of bat biology.

Supported by Grants from Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and from Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP grant 97/02731-8) to A. P. Cruz-Neto. A. S. Abe was supported by a FAPESP grant (1994/6026-9).