Association mapping is rapidly becoming an important method to explore the genetic architecture of complex traits in plants. Over the past decades, a large amount of accessions has been collected for Arabidopsis thaliana. Ultra-high density genotyping followed by careful family-structure analysis has resulted in the assembly of a corepopulation consisting of 360 accessions. In this study we used this population to quantify seed size, germination on water, as well as the germination response to salt, heat and ABA. Experiments were replicated with seeds harvested in two successive years and a high level of heritable variation was observed. Five new natural seed coat mucilage mutants were discovered by analyzing the correlation between dry and imbibed seed size. Interesting correlations between the measured phenotypes and latitude or longitude positions were found. Such local adaptation is the most compelling source of evidence for natural selection during evolution. The combination of highly heritable phenotype measures and an optimally designed population increases the probability to find statistically significant SNP associations. However, no significant SNPs were detected when applying a Bonferroni multiple testing correction, which complicates the discovery of true associations. Due to the importance of seed germination in a plant’s life cycle, a robust biological system is needed in which many loci with small additive effects may determine the final output. Thus, genome wide association can easily be underpowered to efficiently detect such relatively small effect loci. A comparison with traditional linkage mapping in a Bay-0 x Sha RIL population was made in an attempt to enforce the discovery of true associations. De novo candidate genes are listed and prioritized using available expression and annotation data.