Acta Zoologica Academiae Scientiarum Hungaricae 44 (1-2), pp. 1-6, 1998

Title: Introduction to the Symposium: putting the morphometric synthesis to work

Authors: Klingenberg, C. P. and F. L. Bookstein

Authors' addresses: Department of Zoology, Duke University, Durham, North Carolina 27708-0325, U.S.A. and
Institute of Gerontology, University of Michigan, Ann Arbor, Michigan 48109-2007, U.S.A., E-mail: fred@brainmap.med.umich.edu

Abstract: The discipline of morphometrics looks back on a history of about a century. Especially in the last two decades, it has undergone tremendous change described by some authors as a revolution (Rohlf & Marcus 1993). In this short time, numerous methodological approaches have been developed by researchers in various fields, especially evolutionary biology, physical anthropology, paleontology, and systematics. Some of these methods have been consolidated into the morphometric synthesis, of which Bookstein (1998) gives a detailed account. This consolidation, mainly in the past few years, has been every bit as remarkable as the momentous change preceding it. Now that the synthesis is established, the emphasis can shift to the application of morphometrics in various biological disciplines. This is the principal focus of the symposium proceedings in this issue of Acta Zoologica Academiae Scientiarum Hungaricae.

The essay by Bookstein (1998) compares the various methodological approaches in morphometrics. While the paper compares the different approaches in an historical perspective, it is much more than just recounting the sequence of how the invention of new methods rendered older ones obsolete. A surprising result of the morphometric synthesis is that some (but not all) methods that we perceived as competitors a short time ago are now complementary parts within the larger framework of the synthesis. Therefore, this re-evaluation from a "post-synthesis" viewpoint serves two purposes: to readers unfamiliar with morphometric methodology, it offers an introduction to the development of ideas in the field, whereas even seasoned morphometricians should find new insights on how familiar techniques relate to each other.

As morphometricians shift their attention from methodology to application, careful attention to acquisition of data becomes ever more important. The papers by Reig (1998) and by Arnqvist and Mårtensson (1998) discuss the problem of measurement error. Reig (1998) investigates the importance of measurement error relative to differences within and between populations and species. His paper also illustrates a way to deal with specimens missing one or more landmarks (for an alternative approach, see Yaroch 1996). The paper by Arnqvist and Mårtensson (1998) reviews various sources of random and systematic error in morphometric measurements, and uses a simple data set to demonstrate repeatability as a relative measure of error. The example presented is limited to a demonstration of the technique because repeatabilities are computed for principal component scores (which are only meaningful for the particular specimens in the sample). Actual analyses will apply the methodology to other statistics, like loadings (coefficients) for principal components or interpretable scores such as canonical variates (in the context of multigroup comparisons). Despite its simplicity, however, the analysis does illustrate a general point: repeatability is low for principal components that account for small portions of variance, indicating that extra attention to measurement error is necessary whenever subtle features of variation are of interest. The paper by Marcus (1998) also briefly discusses measurement error, including the differences between measurements made by different observers. How serious is measurement error for "real-life" morphometric studies? The answer depends on the context. Studies of variation among different species or across ontogenetic stages will often find a sufficiently large amount of shape variation that the noise introduced by measurement error will have little effect on the results. Studies of intraspecific variation within a particular growth stage, however, will be much more vulnerable because they are aimed at more subtle shape variation. For example, fluctuating asymmetry is a field where the true variation is often not much more than the precision of the measurements. Thus, as asymmetry studies begin to use the methods of geometric morphometrics (Bookstein 1991 pp. 267 ff., Auffray et al. 1996, Smith et al. 1997, Arnqvist et al. 1997, Klingenberg & McIntyre 1998, Klingenberg et al. 1998), repeat measurements will be imperative, just as they are for the conventional measurements of interlandmark distances (Palmer 1994). Allometry is another topic that is equally relevant in the framework of the morphometric synthesis as it has been in more traditional morphometric methods (e.g. Bookstein 1989, Klingenberg 1996, 1998). The paper by Baylac and Penin (1998) uses the traditional analyses of interlandmark distances and the methods of geometric morphometrics to study static allometry in the wings of the fly Drosophila simulans. This study system is particularly interesting because so much is known about the development and genetics of the closely related Drosophila melanogaster. Baylac and Penin (1998) use this source of information to ask whether developmentally distinct compartments of the wing are also different in terms of their morphometric variation, and to interpret the findings of their analyses. For the most part, morphometric approaches have been used in evolutionary biology. The article by Marcus (1998) gives a detailed account of evolutionary changes in fossils of a bovid species that lived on the island of Mallorca in the Pleistocene and early Holocene. This species has a number of morphological features in its skull and jaws that make it truly remarkable. Marcus discusses his morphometric studies of these and other features in the light of functional and ecological considerations. Whereas the paper by Marcus (1998) concerns the evolution over time in what was presumably a single evolutionary lineage, Corti et al. (1998) focus on a group of related species of rodents that currently inhabit different areas in the northern part of South America. Corti et al. (1998) relate morphometric variation to a dendrogram derived from genetic differences among these populations. They used an averaging method for mapping of morphometric variables onto the dendrogram (Rohlf 1997), which is similar to the parsimony methods routinely used in comparative studies of quantitative traits (Harvey & Pagel 1991, Maddison & Maddison 1992; for discussion in the context of morphometrics, see Klingenberg & Ekau 1996). Recently, methods for maximum-likelihood estimation of ancestral trait values have been introduced (e.g. Schluter et al. 1997); this approach, based on an explicit model of evolutionary change (e.g. random walks; Bookstein 1988), will provide the opportunity to integrate morphometrics, evolutionary quantitative genetics, and phylogenetic analysis. While we see promise in this endeavour to map morphometric variation onto a known phylogeny (which will also provide measures of uncertainty for estimates of ancestral morphology), we are pessimistic about attempts to do the reverse, i.e., to estimate phylogenies from morphometric data. The papers by Loy and Capanna (1998) and by Rácz and Demeter (1998) deal with character displacement in related species of moles and shrews, respectively. They are thus concerned with the interactions within and between populations, and with the relationships of populations to their environment. In addition to documenting character displacement, these contributions also underscore that morphometrics has long been, and continues to be, a powerful method for delimiting species, and is thus an indispensable tool in alpha taxonomy. The application of morphometric approaches is not limited to basic science. For instance, the primary example in the essay by Bookstein (1998) is from medicine. As another example from applied science, Rinderer (1998) gives a review of the long-term program to identify Africanized honey bees, and its implementation in a regulatory context. This program, based on samples from thousands of bee colonies, is certainly one of the largest morphometric projects ever realized. Because the methods need to be applied in the context of quarantine measures, it is important that the methods are fast, perform reliably under varied circumstances, and can be used by personnel with little previous experience. Moreover, cost is an additional consideration for such a program. Rinderer reports that the morphometric approach compares favorably to alternatives like biochemical and molecular methods with regard to all these criteria. Altogether, the contributions in this issue encompass most of the range of topics to which morphometric methods have been applied. Where do we go from here? Bookstein (1998) lists a number of further developments in methodology that will add to the flexibility available in the morphometric synthesis. But we also expect a substantial expansion in the range of questions addressed with morphometric methods. Morphometrics has been used mostly in evolutionary biology, but the quantitative analysis of morphology is relevant not only within this field. We see many promising opportunities for morphometric analyses in a number of other disciplines, especially in developmental biology. There has been a recent synthesis of evolutionary and developmental biology (e.g. Raff 1996), which has mostly been a merger of the experimental methodology of developmental biology with a phylogenetic framework of evolution. Using morphometric methods in developmental contexts not only contributes to the unification of developmental and evolutionary biology, but it can yield entirely new results concerning developmental mechanisms. For instance, studies using geometric morphometrics found that the wings of the honey bee (Smith et al. 1997) and of three species of flies including Drosophila melanogaster (Klingenberg et al. 1998) consistently show directional asymmetry between left and right body sides. This implies that a left-right axis must be involved in wing development, which establishes the difference between the two body sides. However, developmental biologists have not yet found a left-right axis in any insect, despite the vast research on specification of body axes in Drosophila, which led some authors to conclude that a left-right axis does not exist in insects (e.g. Raff 1996, pp. 80 f., 302 f.). The morphometric studies clearly established that it does exist (for detailed discussion, see Klingenberg et al. 1998). We expect that morphometric methods will lead more discoveries of morphological patterns that have direct implications for fields such as development and physiology, and therefore can pinpoint new questions for experimental research. Bookstein (1998) ends his article with a general reflection on the origin of quantitative methodology in the sciences. He concludes that it is most effective to use general principles of applied mathematics as the basis for new methodology. The morphometric synthesis is an excellent example of this approach. The next challenge is to integrate these methods into the various disciplines that apply morphometrics, in order to create a unified framework for the quantitative study of morphology. The contributions in this issue are testimony to the fact that this second stage of the synthesis is well underway.

Acta Zoologica Academiae Scientiarum Hungaricae 44 (1-2), pp. 7-59, 1998

Title: A hundred years of morphometrics

Author: F. L. Bookstein

Author's address: Institute of Gerontology, The University of Michigan, Ann Arbor, Michigan 48109, U.S.A.
E-mail: fred@brainmap.med.umich.edu

This paper recounts the development of methods for quantifying, summarizing, and contrasting the spatial relationships of biologically corresponding loci across samples of specimens: the history of morphometrics in the broadest technical sense. In its contemporary version, this toolkit fuses two once-disparate scientific methodologies (Bookstein 1993). The older of the two, the elaboration of diagrams that convey geometrical differences among biological forms to the viewing eye, has been with us (as the art of caricature) since the Renaissance. The junior partner, comprising the strategies of multivariate statistical analysis for optimal summaries of multiple empirical measurements, began to emerge around 1900 in the hands of the biometrician Karl Pearson. This lecture reviews the main themes of their successful combination over the present century.

Acta Zoologica Academiae Scientiarum Hungaricae 44 (1-2), pp. 61-72, 1998

Title: 3D digitizing precision and sources of error in the geometric analysis of weasel skulls

Author: S. Reig

Author's address: Museo Nacional de Ciencias Naturales, CSIC, José Gutiérrez Abascal 2, Madrid 28006, Spain
Mustela@Pinar1.csic.es

Abstract: A collection of 112 weasel skulls was used to investigate various sources of error involved in the morphometric analysis of 3D landmark data, using a Polhemus 3Space digitizer. Average precision in the location of 38 landmarks in three repeated digitizations was 0.303 mm, ranging from 0.204 to 0.542 mm. The Generalized Least Squares fit performed for alignment seems to have little impact on the estimation of digitizing error, absolute deviations being always below 0.01 mm. Merging dorsal and ventral configurations into a complete reconstruction of the skull adds an error ranging from 0.002 to 0.257 mm, which may represent up to a 55.5% increase in digitizing error. Estimation of missing coordinates by multiple regression produced an increase in error ranging from 0.215 to 1.016 mm. Relative to population variability, the amount of digitizing error could represent more than 50% of landmark dispersion when the lowest level of variation (between populations of the same species) was considered. Despite digitizing error, statistical analysis of morphological differences within or between groups showed a good degree of resolution. However, the magnitude of the error shown by some of the landmarks might be too large for some studies of within population or within individual variability.

Key words: geometric morphometrics, 3D digitizing error, missing data

Acta Zoologica Academiae Scientiarum Hungaricae 44 (1-2), pp. 73-96, 1998

Title: Measurement error in geometric morphometrics: empirical strategies to assess and reduce its impact on measures of shape

Authors: Arnqvist, G. and T. Mårtensson

Authors' address: Department of Animal Ecology, University of Umeå, S-901 87 Umeå, Sweden
E-mail: Goran.Arnqvist@animecol.umu.se

Abstract: Random measurement error is ubiquitous in morphometric data, and it can cause serious statistical problems. We stress that measurement error is a potential problem primarily when true phenotypic variation in shape is relatively small, such as in studies of intraspecific variation in shape. A model for the partitioning of measurement error in landmark based morphometrics is presented. The impact of measurement error can be reduced in a number of ways, depending on the methods used to collect, process and analyse data, and we give some practical advice. We also recommend that repeated measures of all individuals are taken routinely in morphometric studies where measurement error may be a potential problem. This enables both a quantification, by estimating repeatabilities from analyses of variance, and a reduction, by averaging repeated measures, of the relative impact of measurement error. We perform an analysis of shape variation in a uniform sample of young perch (Perca fluviatilis), solely aimed at illustrating how different components of measurement error can be quantified, and demonstrate (a) that estimates of repeatability will only be informative of the error components that are actually repeated in each repeated measure, (b) that the relative impact of different components of measurement error can be partitioned and assessed by planned hierarchical repeated measurement protocols followed by nested analyses of variance, (c) that measurement error is unevenly distributed among different shape variables and (d) that the relative magnitude of ME in a given shape variable can be reduced to an estimable extent by averaging several repeated measures.

Key words: shape analysis, repeatability, repeated measures, measurement error, Perca fluviatilis

Acta Zoologica Academiae Scientiarum Hungaricae 44 (1-2), pp. 97-112, 1998

Title: Wing static allometry in Drosophila simulans males (Diptera, Drosophilidae) and its relationships with developmental compartments

Authors: M. Baylac and X. Penin

Authors' address: Muséum national d'Histoire Naturelle, Groupe de travail Morphométrie et Analyse de formes, Laboratoire d'Entomologie (EP90), 45, rue Buffon F-75005 Paris, France
E-mail: baylac@mnhn.fr

Abstract: Static allometric patterns of the wing of Drosophila simulans males were estimated by multivariate regression of GLS procrustes residuals on the log of centroid-size, by thin-plate splines regression visualisations, and by multivariate allometry of inter-landmark distances. The allometric shape component equals 6% of the total shape variance. Allometric patterns involve mainly the distal and central parts of the wing. They correspond to a relative contraction of the distal part of the wing, from the posterior cross-veins to the apex. The proximal region of the wing undergoes the lowest allometric transformations. Relationships between allometry and developmental compartments were investigated for uniform transformations. Uniform allometric patterns are significant only for the anterior wing compartment. Principal axes are congruent with known directional patterns of cells clones and landmarks. On the whole, and whatever the compartments are, allometric changes appear basically of the non-uniform type.

Key words: static allometry, multivariate allometry, Procrustes superimposition, developmental compartments, uniform transformations

Acta Zoologica Academiae Scientiarum Hungaricae 44 (1-2), pp. 113-137, 1998

Title: Variation in selected skeletal elements of the fossil remains of Myotragus balearicus,
a Pleistocene bovid from Mallorca

Author: L. F. Marcus

Author's address: Department of Biology, Queens College of the City University of New York
and Department of Invertebrates, American Museum of Natural History
E-mail: lamqc@cunyvm.cuny.edu

Abstract: Three dimensional landmark morphometrics for skulls, mandibles, and metacarpals were used to examine variation in the island Pleistocene bovid Myotragus balearicus from three cave sites in Mallorca. Traditional caliper distances were studied for earlier data collected on metatarsals of the same species from only one cave. Missing data is a major problem for fossil skulls, and mandibles and a large number of specimens and landmarks were ignored to have a full data set for multivariate analysis of a restricted sample of specimens. More compact bones like metacarpals have fewer problems. The specimens are highly variable with most coefficients of variation larger than 7 even for samples presumably from one time level in one cave. This study corroborates Hamilton's (1984) finding that the lower levels in the cave have larger individuals and documents a significant difference over a range of about 11000 years in size of metacarpals from Son Muleta cave. The animal became smaller in the early Holocene before becoming extinct. The other two caves, Cova del Moro and Son Maiol, have different mean sizes, the former similar to the upper level at Son Muleta, while the latter is similar to the lower level. Patterns of variation and the problem of recognizing sexes in these samples are discussed.

Key words: geometric morphometrics, Bovidae, multivariate analysis, three dimensional data, Pleistocene

Acta Zoologica Academiae Scientiarum Hungaricae 44 (1-2), pp. 139-150, 1998

Title: Phylogeny and size and shape changes in the skull of the South American rodent Proechimys

Authors: Corti, M., M. Aguilera+ and E. Capanna

Authors' addresses: Dipartimento di Biologia Animale e dell'Uomo, Universitá di Roma 'La Sapienza', via Borelli 50, I-00161, Roma, Italia, E-mail: corti@axrma.uniroma1.it
+Departamento Estudios Ambientales, Universidad Simón Bolívar, Apartado 89000, Caracas, 1080-A, Venezuela

Abstract: The rodent genus Proechimys has undergone rapid chromosomal speciation during late Pleistocene, resulting in several species and subspecies occurring in Venezuela. Here we present hypotheses for patterns of morphometric change in the skull based upon genetic relationships between two species of the "guairae"ª complex (P. guairae and Proechimys sp.) and P. trinitatis. Morphometrics of these species was studied using a geometric approach, and shape changes were visualised with respect to the branching in the allozymes tree. There is a consistency between morphometric change in the dorsal side of the skull and genetic variation, suggesting a phylogenetic explanation; however, there is incongruency between genetics and morphometrics of the ventral side of the skull. Incongruity may result from mutation rates which vary for both allozymes and morphometry and from the evolutionary forces which may act differently on the functional units (auditory, feeding, respiratory, olfactory) of the ventral side of the skull.

Key words: phylogeny, geometric morphometrics, rodents, Proechimys

Acta Zoologica Academiae Scientiarum Hungaricae 44 (1-2), pp. 151-164, 1998

Title: A parapatric contact area between two species of moles: character displacement investigated through the geometric morphometrics of skull

Authors: Loy, A. and E. Capanna

Authors' address: Dipartimento di Biologia Animale e dell'Uomo, Universitá di Roma 'La Sapienza', 50 Via Borelli, 00161 Rome, Italy, E-mail: annaloy@caspur.it

Abstract: Character displacement is investigated in sympatric populations of two related species of European moles, Talpa europaea and T. romana, along the zone of parapatric contact in Central Italy. Morphological variation of the two species is analysed through the geometric morphometrics of skulls by comparing sympatric to allopatric populations. The morphometric approach enabled us to consider separately the effect of size and shape on character displacement, while the analysis of relative warps allowed us to recognise and visualise specific regions of the skull involved in the phenomenon. Results suggest that size and shape can play a role in the local differentiation of the two species. Differences in size are sharper in the sympatric populations, while shape differences possibly related to character displacement are expressed by few relative warps in both the dorsal and the ventral view of the skull. Morphological features involved are concentrated in the zygomatic region, thus suggesting that character displacement is mainly related to the efficiency of food assumption and processing.

Key words: character displacement, geometric morphometrics, Talpa, Italy

Acta Zoologica Academiae Scientiarum Hungaricae 44 (1-2), pp. 165-175, 1998

Title: Character displacement in mandible shape and size in two species of water shrews (Neomys, Mammalia: Insectivora)

Authors: Rácz, G. and A. Demeter

Authors' addresses: Department of Zoology, Hungarian Natural History Museum
Budapest, Baross u. 13, H-1088 Hungary

Present addresses: Department of Biology, Castetter Hall, University of New Mexico, Albuquerque, NM 87131, U.S.A. E-mail: raczg@unm.edu
Department of Nature Conservation, Authority for Nature Conservation, Ministry of the Environment, Budapest, Költõ u. 21, H-1121 Hungary

Abstract: Landmark coordinates were digitized from images of mandibles of trapped specimens of two species of water shrews (Neomys fodiens and Neomys anomalus, Mammalia: Insectivora), and unidentified mandibles from owl pellets collected in Hungary. Discriminant analyses based on variables describing size, affine and nonaffine shape changes were carried out using positively identified specimens as training sets. The results indicate that the two species can be separated using both size and shape variables, but the best discrimination can be achieved by using both set of variables.
The two species were found to be more different at localities where they occur sympatrically. The more abundant Neomys anomalus is more stable in its characters, while the less abundant species, Neomys fodiens, appears to lessen competition by shifting its characters at localities where both species co-occur.

Key words: Neomys, water shrews, geometric morphometrics, landmarks, character displacement

Acta Zoologica Academiae Scientiarum Hungaricae 44 (1-2), pp. 177-194, 1998

Title: The identification of africanized honey bees: an assessment of morphometric, biochemical, and molecular approaches

Author: T. E. Rinderer

Author's address: USDA-ARS Honey-Bee Breeding, Genetics & Physiology Laboratory, 1157 Ben Hur Road, Baton Rouge, Louisiana 70820, U.S.A. E-mail: trindere@asrr.arsusda.gov

Abstract: African honey bees (Apis mellifera scutellata) were introduced into Brazil in 1955 with the intention of providing improved honey bee breeding stock for Brazilian apiculture. The spread of the descendants of the introduced African bees, known as Africanized bees, was a matter for scientific study and regulatory concern. This attention produced the need for an identification tool that could be employed in research, survey, and detection and regulation. In part due to a long history of study and in part due to its intrinsic value, the discriminant analysis of morphometric data has become the tool of choice for identifying Africanized honey bees. Cost of analysis led to the development of simple methods to screen large numbers of samples without sacrificing the overall quality of identifications. With these screening procedures, all colonies that are determined to be European at a P >= 0.99 are considered European. All colonies that are not determined to be European are considered unidentified. Samples which remain unidentified after the initial screening are identified by a more complex morphometric procedure called USDA-ID. USDA-ID was developed primarily to provide several laboratories that were established to morphologically identify honey bees for regulatory purposes more accurate identification tools based on new discriminant analysis procedures. The characteristics of these procedures are discussed and their weaknesses and strengths are compared to those of several other identification tools.

Key words: Apis mellifera scutellata, Africanized bees, morphometrics, biochemical identification, molecular identification