SPECIATION OF THE TINTINNID GENUS CYMATOCYLIS BY MORPHOMETRIC ANALYSIS OF THE LORICAE

Samples of the tintinnid genus Cymatocylis were collected at an oceani c site near South Georgia in January 1990. The shapes and sizes of lor icae observed included most forms previously reported by other authors and were representative of the entire genius. Measurements were taken from the loricae of over 700 specimens and 201 photomicrographs were obtained, from which further detailed measurements were taken. Univari ate frequency histograms and bivariate scatter plots of the morphometr ic measurements were compared with multivariate techniques including: hierarchical nearest neighbour cluster analysis, linear discriminant a nalysis and canonical analysis with resubstitution on the model to 95% confidence intervals. Fourier transforms of digitised images of the p hotomicrographs were utilised as functions of the overall shape of the organisms, and input to both the linear discriminant function and can onical function with resubstitution on the model to 99% confidence int ervals for comparison with results obtained from the manual morphometr ic measurements. Linear discriminant analysis showed 5 clear taxonomic classes corresponding to the original descriptions of C. calyciformis , C. convallaria, C. vanhoffeni, C. parva and C. drygalskii. Resubstit ution onto the canonical models gave correct classification for the ma nual morphometric data and 100% correct classification for the Fourier transform data. These results showed that a clearer discrimination wa s obtained by utilising a multivariate 'description' of the overall sh ape. The cluster analysis showed that absolute size was not necessary for the identification. The univariate and bivariate approaches demons trated some discernible separation, but with considerable overlap betw een species, especially C. vanhoffeni and C. drygalskii. These statist ical methods were used to demonstrate that clear discrimination can be obtained from morphometric data and should allow for the development of automated taxonomic classification.