Phenotypic variation is the raw material of evolution. Standing variation can facilitate response to selection along “lines of least evolutionary resistance”, but selection itself might alter the structure of the variance. Shape was quantified using 2D geometric morphometrics in Palmatolepis conodonts through the Late Devonian period. Patterns of variance were characterized along the record by the variance-covariance matrix (P-matrix) and its first axis (Pmax). The Late Frasnian was marked by environmental oscillations culminating with the Frasnian/Famennian mass extinction. A shape response was associated with these fluctuations, together with a deflection of the Pmax and the P-matrix. Thereafter, along the Famennian, Palmatolepis mean shape shifted from broad elements with a large platform to slender elements devoid of platform. This shift in shape was associated with a reorientation of Pmax and the P-matrix, due to profound changes in the functioning of the elements selecting for new types of variants. Both cases provide empirical evidences that moving adaptive optimum can reorient phenotypic variation, boosting response to environmental changes. On such time scales, the question seems thus not to be whether the P-matrix is stable, but how it is varying in response to changes in selection regimes and shifts in adaptive optimum.

Platform elements of the conodont genus Palmatolepis have been collected in seven Frasnian and twelve Famennian levels of the Col des Tribes section (Montagne Noire, France). Age was estimated for each level using a biostratigraphic analysis (Girard et al. 2014). The study focused on four well-represented subgenera: Manticolepis, Panderolepis, Deflectolepis and rhomboidea. 2D pictures of the elements were taken, the oral surface facing the camera. The outline of the element was described with 64 equidistant points. Based from the corresponding coordinates, radii (e.g. distance from each point to the centroid) were calculated and analysed using a Fourier-based morphometric method. The coefficients of nine harmonics were retained, thus delivering 18 shape variables (coefficients A1 to B9). These coefficients have been standardized by size, estimated by the zeroth harmonic (A0). A total of 1535 elements have been included in the analysis. 

Renaud_Girard_Dufour_2021_Rcode.R

This file contains the script of the multivariate analyses used in the study.

Palm_CTselect_2021.txt

This file contains the dataset. Successive columns correspond to the following variables. Section: CT outcrop; level: numbering of the level along the section; depth: depth in the section in cm; agemodel: estimated absolute age of the level according to Girard et al. (2020) updated after Becker et al. (2020); LevSide: code combining level, subgenus and side; Zone: biostratigraphic zones, numbered according to their temporal order; Side: right/left side; collage and photo: person responsible for orienting and picturing the elements (CG: Catherine Girard); species: tentative specific identification; Subgenus: subgeneric attribution; ssgleve: code combining subgenus and level; A0: zeroth harmonic (size); A1 to B9: Fourier coefficients (shape variables).