Report on Paris Workshop (Paris Atelier)

October 26 to October 31 1992

No other place would have been more prestigious to host the Paris Morphometric Workshop than the Museum National d'Histoire Naturelle: Much of the modem Evolutionary and Systematic thought has its roots in the buildings at the Jardin des Plantes, between rue Buffon and rue Cuvier. The Morphometric workshop, the fourth in order after Ann Arbor, Stony Brook, and Valsain, was held at the Salle Lamark from October 26 to October 31 1992, organized by Michel Baylac and the Groupe de travail Morphometrie et Analyse de Forme. First day plenary lectures introducing traditional and new morphometrics took place at the Amphiteatre du Laboratoire de Paleontologie, at the entrance of the Comparative Anatomy Hall, an exquisite preface to Geometric Morphometrics, that many of us consider an extension of this old science. The morning program was moderated with authority by L.F. Marcus and included an opening lecture by R. A. Reyment, "Morphometrie multivatiee: aspects historique", which was presented in French, followed by Fred Bookstein with "Landmarks and the measurement of deformation", and by Jim Rohlfs "Seeing shape change" (the last two in English). The audience was led through the history of morphometrics, from the original roots of Huxley's concept of differential growth and the original intuition of D'Arcy Thompson on the need to interpret shape changes geometrically, through the coevolution - if this word is allowed here - of morphometrics with multivariate statistics, the multivariate morphometrics, to the emerging and promising geometric morphometrics, a full grown language to describe shape and size changes. The other days were spent at Salle Lamark, with morning and afternoon lectures, and hands on computers to experiment with software on data. All was directed by our marvellous three, Fred Bookstein, Les Marcus and Jim Rohlf, with the authoritative presence of Richard Reyment, a corroborated attack group that has now survived four morphometrics workshops and that, apparently, is ready for bigger enterprises: per aspera ad astra (check morphmet!!). There were seventeen students in Paris, representing four European countries, France, Italy, Poland and Spain.

What news to report? In Paris there was no presentation of hardware equipment for image and data acquisition, following the suggestion from the Valsain workshop. All attention of participants was devoted to the understanding of the concepts and the methods of geometric morphometrics, as well as to other more traditional aspects, such as the historical 'size and shape' problem of multivariate morphometrics, which was deeply discussed by R. Reyment and F. Bookstein. But many hours were spent with hands on the PCs. The real news is that all of geometric morphometrics - shape or Bookstein coordinates, Procrustes and Thin-Plate Spline methods were explained at their best. For those who participated in previous workshops, things that required days (even years) to understand, such as Warps - principal, partial, relative - were digested by participants without any apparent rejection. This is undoubtedly a merit of the teachers that have accumulated a considerable experience during the previous workshops. However, superimposition methods have been under development - and they still are - for very few years. There is some more experience in their use now, so that their teaching and understanding them becomes progressively more comfortable. The writer participated in the two-weeks Stony Brook 1990 workshop, and would liked it to have been brought into geometric morphometrics as easily as participants could do during the 5 days 1992 Paris workshop! A testimonial of the 'experiment and develop' side of this science was offered by Jim Rohlf's TPSRW program, which computes relative warps with a variety of options, and that had successive releases during the workshop, even 2 within an hour. There was time for presentations of applications and new techniques. We had an important lecture by Jim Rohlf concerning MANOVA and CVA on partial warp scores, with a demonstration of a real analysis on moles; projections of individuals onto CV axes were visualized in terms of splines by the new software developed for this purpose (see after), so that now we have a new way to look at statistics.

This was proceeded and followed by two presentations of applications. Anna Loy presented her work with L.F. Marcus and M. Corti on the systematics of Old World moles (Talpidae: Insectivora), a complete landmark analysis of size and shape of the skull of 7 species, including shape coordinates and relative warps (the complete paper is going to appear in the book with the proceedings of Stony Brook and Valsain workshops); Jean-Pierre Hugot gave a talk on his research with Michel Baylac on the coevolution of helminth parasites with primates and other mammals, based upon a collection of landmarks recorded on the buccal apparatus. There were two official software first presentations: GRF-ND, or 'tumbling turtles', a beautiful program by Dennis Slice (we missed him and his announcements in Paris!) that computes Procrustes analysis in two or three dimensions; a real improvement of the 'old' GRF by Rohlf and Slice, with more analyses and options and a real output. The second software was TPSRW-REG, 'Thin Plate Spline Relative Warps Regression", by Jim Rohlf, that relates Relative Warps to external measures of variation, such as for example latitude, longitude, altitude, ecological parameters, etc. that may help in explaining the warping biological phenomenon under study; or, relative warps and splines can be expressed alternatively as the effect of a multivariate analysis of partial warps. Finally, Les Marcus gave an introductory demo of MATLAB, a programmable package performing matrix algebra easily; some are passing to MATLAB, apparently abandoning other programs. There were two important points which came out during the workshop. First, there was a need for stand-alone programs. In particular, to compute shape coordinates, centroid and baseline size, and the affine component of shape variation from a given set of landmarks, one has to use UNICOORD.SAS which is written in the IML language of SAS by Les Marcus (the listing is published it the appendix of the 'Black' book of Reyment). There are two limitations that one has to face to get these transformations and results: SAS is a big and expensive package, and moreover does not allow in IML 'big' matrices that are usual in a morphometric study. Les Marcus promised to write a program in QBASIC to solve these problem, and there is a beta release now. Secondly, but most important, there was a debate during the last day, but that should have lasted for more time, concerning the amount of knowledge of matrix algebra and manipulation that one must possess. There were two extreme positions: it would be enough to have a general comprehension of the mechanisms at the base of the numerical methodologies (with main allusion to TPS and allies) to face a morphometric study, or should one have a thorough knowledge of each single step in one's given analysis? The question is fundamental to many, and perhaps others should give their opinion on morphmet. Some of the participants, and obviously the teachers, have been in more than one morphometric workshop after Stony Brook. The Paris morphometrics workshop has also been an occasion to verify ideas, make clear things that were obscure, keep in touch with the day by day appearing news in morphometrics, see the work of others and have a reaction to the work done; why not a real meeting sometime?

Marco Corti

Dipartimento di Biologia Anmale e dell'Uomo Universita' di Roma "La Sapienza' Via Borelli 50, 00161

December 21, 1992