Supplementary data for: Another worm bites the dust

Based on a dataset of most mid-Paleozoic scolecodonts photographed in the literature, the Carnegie Museum of Natural History, and newly presented data from the Appalachian Basin, we find that scolecodonts (polychaete jaw elements) decrease in size across the Late Devonian boundary, then increase again to baseline in the Carboniferous. The majority of the data representing small scolecodonts during the extinction event are newly presented data in this study. These individuals were sampled from black shales in a small geographic range at high stratigraphic resolution. Even when excluding these new data, scolecodonts after the Frasnian-Famennian boundary are smaller than scolecodonts before the boundary. Lithology or water depth has a relationship with scolecodont size, but those factors alone cannot explain the size reduction observed across the extinction (i.e., we also find big scolecodonts in black shales). This represents a novel, previously-unreported pattern in scolecodont size occurring across the Late Devonian mass extinction. We interpret this as an oxygen stress-driven occurrence of the Lilliput Effect. The biological mechanism driving the change appears to be adaptation within taxa–either evolutionary or epigenetic–not a dying off of large clades. Although oxygen stress is known to reduce biomass of modern communities of polychaetes, this is new evidence of size reduction due to oxygen stress.

Methods
To increase our sample size of Appalachian Basin Lower Paleozoic scolecodonts, we also measured the scolecodonts from the extensive collections in the Carnegie Museum of Natural History (CMNH). These scolecodonts were collected, prepared, and described by former curator of the Museum E.R. Eller in the mid-1900s, and have also been described by M. Eriksson and C. Bergman. Some of these were photomicrographed under 40x magnification at Williams College, while most were photographed at the Museum using a ProScope EDU 300 Portable Standalone Digital Microscope. 

Literature review
We conducted an extensive literature review to collect data on as many published images of Devonian and Carboniferous age scolecodonts as possible, as well as a large sampling of Silurian scolecodonts. This search was conducted primarily using Google Scholar. We excluded from the dataset some images that were unusable because the scolecodonts were obscured or images were too low-quality; stratigraphic resolution was too poor; scale bars were absent and magnification could not be reliably determined; images were sketches rather than photomicrographs; or scolecodont affinity was uncertain. We also searched the literature for redox proxy data (i.e., molybdenum and total organic carbon) from the same strata that scolecodonts were sampled from, but availability of these data were limited. Scolecodonts in our literature search represent 73 localities where no one locality makes up more than 9.7% of individuals from the literature (Figure 8). We examined 711 scolecodonts from the literature and 77 scolecodonts from our own samples, thus our scolecodonts made up 10.8% of the study. 

Measurement
The magnification of photomicrographs taken at Williams College varied but was recorded and calibrated to a scale in ImageJ to enable precise measurements. Photomicrographs taken at the Carnegie Museum were calibrated to a scale bar and directly measured using the ProScope software. Photomicrographs obtained from the literature were uploaded into ImageJ, calibrated using the scale bar provided in the publication, and measured in ImageJ (Figure 9). Sizes recorded by authors in the literature were ignored to ensure consistency of measuring methodology.

The length of a scolecodont is the greatest dimension of a jaw approximately parallel to the dentary, while the width is the greatest dimension of a jaw roughly perpendicular to the length (Jansonius and Craig 1971). 

Statistical analysis
Statistical analyses were performed in RStudio using the tidyverse, maps, mapdata, rstatix, ggpubr, dplyr, RCurl, cowplot, reshape2, scales, plyr, and RColorBrewer packages. Initial plots detected some extreme outliers: scolecodonts from Szaniawski and Wrona (1973) were extremely large compared with the rest of the dataset. These individuals were excluded from all other analyses because the images in this paper are sketches, not photomicrographs, and therefore scale can not be reliably determined. Likewise, the individual pictured in Eller (1936) was excluded because it is a sketch. Scolecodonts from other Eller papers were excluded because the exact same individuals were measured in-person at the CMNH, or because they were sketches and therefore deemed unreliable. A few CMNH scolecodonts whose age could not be determined were also excluded from all further analyses. When a scolecodont was partially obscured in one dimension, the obscured dimension was not measured. Rows without a value assigned for the variable of interest were excluded for that analysis but included in other analyses. 

To ensure replicability of results, we recorded the figure number and DOI of every image retrieved through literature review. Where possible, every individual was assigned to a lithostratigraphic age and epoch as well as basin, unit, and locality, based on information provided in the text and cross-referenced with other publications and shapefiles. For individuals from the Late Devonian, a separate variable specified whether they were dated to before, during, or after the Frasnian-Famennian boundary. Decimal degree latitude and longitude were sourced directly from the text or estimated using Google Maps. Taxonomic and morphologic (component) information were sourced from the text when possible and aggregated to the family level using the World Register of Marine Species. Though Late Devonian paleogeography is a work in progress, we loosely grouped individuals by province and paleocontinent based on basin and information provided in the text. We also recorded lithologic information, redox proxies where available, and depth of core samples. We also attempted to identify the species and component of the scolecodonts from the samples we collected, though we acknowledge that we lack expertise in polychaete jaw taxonomy.