In the Acquisition and Processing sections, [#] refers to the references cited in the publication.

Sampling and study population
Barnacles were collected from one 20-m-long rocky outcrop, approximately near Punta Gaspareno, Baja California, Mexico (23°10'58.09 oN, 110° 8'26.51 oW) in October 2011. Aggregations of barnacles at seven different densities--1, 2, 4, 12, 17, 22, and 44 individuals per 10 cm2-- were sampled haphazardly using a 10 cm2 quadrat (see Table 1). We found few individuals brooding embryos in the lowest density groups; only one of eight groups with a density of two individuals per10 cm2 and none of the four solitary individuals had embryos. Lower densities (1 or 2 individuals) were sampled more than once in an attempt to find individuals that were brooding embryos for paternity analysis. In total, 119 barnacles were sampled, but of these only 21 individuals exhibited brooding embryos, 14 of which yielded larvae for paternity analyses (Table 1; larval culturing description below). Individual barnacles were scraped off of rocks with 1 mm-thick metal paint scrapers and care was taken minimize damage to the bottom of the peduncle so that adults were brought back alive. Animals from each aggregation were then placed in individual zip-lock bags and kept moist with rinses of fresh seawater every 6-8 hours until they arrived at the laboratory ~24 hours later.

Larval culturing and tissue sampling
Prior to extraction of the disc-shaped embryo sacs (lamellae), each individual’s peduncle was dipped in a 10% bleach solution, the capitulum was rinsed in 90% ethanol, and then the whole animal was rinsed thoroughly with 35 ppt artificial seawater (ASW, Instant Ocean). Lamellae (both discs) were extracted from the mantle cavity with forceps, rinsed with ASW, and transferred to clean, 100 ul plastic beakers containing 50 ul of fresh ASW treated with 1 mg/l each of streptomycin and penicillin to limit bacterial and fungal growth [71]. Larvae from each brood were reared in separate cultures in the dark at 25°C [71]; each culture was checked daily for hatching. After hatching, swimming larvae were transferred to new beakers and fed with Rhodomonas salinas and Isochrysis galbana at concentrations of 10,000 cells ml-1 each for ~48 h until sufficient numbers of stage-II larvae could be collected (50 or more) and preserved in 70% ethanol. Larvae were fed for 48 h after hatching because larger larvae yielded more DNA, facilitating individual genotyping. Peduncle tissue of mothers with broods that had 50 or more stage-II larvae were sampled and preserved in ethanol.

DNA was extracted from ~25 mg of adult muscle tissue using a modified CTAB protocol (Doyle and Doyle 1987) with two chloroform/isoamyl isolations and two 70% EtOH washes. Precipitated adult DNA was re-suspended with Qiagen EB buffer (10 mM Tris, ph 8.5; Qiagen Valencia CA). DNA from larvae were extracted individually in 200ul 96-well PCR plates using 25 ul of extraction buffer comprised of 0.5% tween, TE buffer (10 mm Tris, 1 mM EDTA), and 2.5 ul 20 mg/ml Proteinase k (Bioline). Larval extractions were incubated at 60°C for 4 hours followed by 30 minutes at 95°C and stored at -20°C. Raw (unprecipitated) larval extractions were used directly in PCR.

Genotyping and paternity analysis
Brooding individuals (mothers) and offspring were genotyped with at least three of five loci: Pole 1, Pole 8, Pole 25, Pole 29, and Pole 44 [72]. These loci have a high number of alleles (5-23) and gene diversity (average expected heterozgosity=0.59), show no evidence of null alleles, and exclusion probabilities calculated from the four most commonly genotyped markers (Pole 1, Pole 8, Pole 25, Pole 44) showed high discriminatory power (0.94; genotype data from Gaspareno, Mexico in [72]). PCR was carried out as described in Plough and Marko [72] and fragment analysis was run on the ABI 3100 sequencer at the Arizona State University DNA Lab. Electropherograms were scored by eye using LIZ600 (Applied Biosystems) as an internal size standard on the Peak Scanner software v. 1.0 (Applied Biosystems). Note that because P. elegans is hermaphroditic, all individuals, including brooding mothers are potential fathers.

Paternity analysis was performed using the program GERUD 2.0 [73]. The software does not allow for missing data, so only larvae that successfully amplified at all loci were included in the analysis of each brood. In one brood (c22), two larvae (0.48% of all genotyped larvae) were each homozygous at a single microsatellite locus for an allele that differed from the mother’s, violating assumptions of Mendelian segregation. These larvae were removed from the analysis as possible contaminants from another brood but de-novo mutation at this locus could also explain the observed segregation pattern (e.g. [74,75]). Multiple genotype array solutions for the fathers were ranked by likelihood using the default test for Mendelian segregation. Though GERUD2.0 has the option of ranking solutions by allele frequencies from a reference or base population, we were not able to use this feature because some of the broods exhibited rare alleles not present in the genotype data from [72]. To determine if offspring were likely sired by fathers in the same physical aggregations, we examined whether the alleles observed in larval broods were also present in the genotypes of adults from the same aggregations in which those broods were collected. We compared larval and adult genotypes from each of two low density aggregations (2 individuals/10 cm2 and 4 individuals/10 cm2) and eight broods from three high density aggregations (17, 22, and 44 individuals/10 cm2; Table 1). All of the adults within the two low-density groups were genotyped, but in the three higher density aggregations we were able to genotype only ~80-95% of the adults because some individuals were damaged during collection and/or transport, compromising the quality of the DNA.

Power analysis of paternity
To determine the power to detect different levels of multiple paternity given the offspring sample size and the number of contributing fathers, we ran simulations using GERUDSIM 2.0. We determined the proportion of simulations out of 1000 that correctly assigned the true number of fathers (from 2-5), given various sample sizes of larvae genotyped (range, two - 96; actual mean sample size across the 14 broods =29.74) and the population allele frequencies of the four most commonly used markers (Pole 1, Pole 8, Pole 25, and Pole 44) from Baja California, Mexico [72]. Simulations were run assuming that the mothers’ genotype was known and that the total offspring number per female (the average number of fertilized eggs) was 5000. In the simulations, five thousand offspring is then split among the true number of fathers for a given simulation scenario (e.g. for the 3 father simulation with equal reproductive contribution, each would be assigned 1666 offspring). Fecundity in P. elegans has not been measured systematically, but appears to range from a few thousand to 10’s of thousands of eggs per female based on observations from this study, and estimates for the related barnacle Pollicipes pollicipes are similar [76]. Simulations with greater than 5000 total offspring ran extremely slowly in GerudSim2.0, but a few trial runs with 25,000 vs 5,000 total offspring yielded similar power results, so we set the total number of offspring in a brood to 5000. We also determined the power to detect the true number of fathers in a brood when there was skew in paternal contribution: 1/5 the contribution from one father and equal contributions from the others. These simulations were run only at the approximate mean offspring sample size (30). For example, in the case of five fathers, we assigned one father only 200 offspring and the other fathers 1,000.