Methods: EXTINCTION AND NATICID PREDATION OF THE BIVALVE CHIONE VON MÜHLFELD IN THE LATE NEOGENE OF FLORIDA

METHODS

Morphometrics

Samples of C. erosa were obtained from the following deposits and localities (Figure 3) (sample sizes in parentheses after sample name, sample number prefixes explained at the end of this paragraph): (A) Jackson Bluff Fm. - UF7103 and UF3037 (25), Leon Co. (B) Pinecrest Beds - UF78788 (27), below Unit 8, Sarasota Co.; R016 (13), below Unit 7, Sarasota Co.; R017 (28), Unit 7, APAC pit, Sarasota Co.; R012 (12), Unit 6, APAC pit, Sarasota Co. (C) Caloosahatchee Fm. - R001 (30), La Belle. Samples of C. cancellata are as follows: (A) Bermont Fm. - R018 (27), Pit 1, Leisey, Sarasota Co.; R019 (36), Pit 3, Leisey, Sarasota Co. (B) Ft. Thompson Fm. - R010 (35), Englewood, Charlotte Co.; UF78815 (40), Englewood, Charlotte Co. (C) Anastasia Fm. - 777T (17), Palm Beach Co. (D) Recent - Lake Worth (30), Palm Beach Co. Sample prefixes are: R - author's collection; UF - Florida Museum of Natural History; T - Tulane Paleontological Collections (samples now at UF).

All samples were analyzed morphometrically using a geometric approach to separate samples of C. erosa and C. cancellata, and also to interpret the morphological differences between the two species. Left valves were imaged using a flatbed scanner (Hewlett Packard ScanJet IIcx), and landmark locations digitized with image analysis software (Mocha version 1.2, Jandel Scientific). All landmarks selected for analysis originate from either the hinge region or the adductor muscle scars (Figure 4), with the exception of the ventral ends of the lunule and escutcheon. These landmarks represent a mixture of Type I and Type II geometric landmarks (Slice et al. 1996) and are identical to those used by Roopnarine (1995) to generate inter-landmark distance measures.

The digitized landmarks were converted to shape coordinates (Bookstein 1991) by a process of translation, rotation and reference to a common baseline. This procedure effectively separates shape and scale information, with specimen size being recorded separately as Centroid Size (Bookstein 1991). Each specimen of all samples was then compared to a Procrustean consensus reference form (Rohlf 1996), and the comparison decomposed into uniform (affine) and non-uniform (non-affine) components using TPSRELW (Rohlf 1997). Uniform and non-uniform transformations are independent descriptors of shape, and represent global affine shape differences and non-affine localizable shape differences respectively. Uniform shape transformations summarize deviation from the consensus reference form as an affine transformation (Bookstein 1991; Slice 1996), decomposing the transformation into stretching and shearing components. Non-uniform transformations ("partial warps") on the other hand describe non-affine localizable shape differences. Distribution of the consensus' landmarks in Kendall shape space (Kendall 1986) is summarized by a series of tangential "principal warps''. The distribution of a specimen on the Euclidean projection of the principal warps is in turn summarized by a series of partial warps, which describe the non-uniform transformation of consensus form to specimen. Non-uniform transformations can be visualized by the now well known thin plate spline diagrams.

Specimen scores on the uniform and non-uniform components were used as shape variables in conventional multivariate analyses of variance (MANOVA) and canonical variates analyses (CVA). There were no significant shape differences among the samples obtained from a single formation, so those samples were combined in all cases (Roopnarine 1995). Static allometry was assessed within each sample by regression of shape scores on centroid size (the only multivariable measure of size that is uncorrelated with shape in the absence of allometry), but no significant allometry was detected within any sample.

Drilling morphometry

Samples of drilled specimens were drawn from those used in the morphometric analysis, but in some cases exceed the sample sizes used for morphometric description. Samples are (sample sizes in parentheses): Pinecrest Beds (89), Caloosahatchee Fm. (11), Bermont Fm. (102), Ft. Thompson Fm. (101), Recent (32). The Anastasia Fm. was excluded because of small sample size, and the results from the Caloosahatchee Fm. should be treated cautiously for the same reason. The frequency of drilling was not measured because the samples were not derived from bulk samples.

Three drilling parameters were documented for the two Chione species: drill hole location, drill hole size, and valve thickness. Prey size was assessed as maximum valve height, and outer drill hole diameter served as a proxy for predator size (Kitchell et al. 1981). All measures were made with digital calipers to the nearest 0.01mm, except valve thickness, which was measured at the ventral margin of the valve with a micrometer screw gauge to the nearest 0.01mm. Hole location was mapped onto a valve using the image techniques and morphometric baseline described above to record the geometric location of the hole. In this respect, the hole centroid is treated as a landmark of the shell. Selectivity and randomness of hole location were tested by dividing the available valve surface into 20 equal area sectors (Figure 5), recording the frequency of drilling in each sector, for each sample, and comparing sample means and variance with likelihood ratio tests (Manly 1986) of means and mean deviations respectively.

Relationships between drill hole diameter and valve height, and valve thickness and height, were formulated using Model II regression of log transformed values. The significance and equality of regression parameters were tested by comparison of bootstrapped percentile confidence intervals based on 1,000 regression iterations (Efron and Tibshirani 1993).