Although there is a heritable basis for many animal behaviors, the genetic architecture of behavioral variation in natural populations remains mostly unknown, particularly in vertebrates. We sought to identify the genetic basis for social affiliation in two populations of threespine sticklebacks (Gasterosteus aculeatus) that differ in their propensity to school. Marine sticklebacks from Japan school strongly whereas benthic sticklebacks from a lake in Canada are more solitary. Here, we expanded on our previous efforts to identify quantitative trait loci (QTL) for differences in schooling tendency. We tested fish multiple times in two assays that test different aspects of schooling tendency: 1) the model school assay, which presents fish with a school of eight model sticklebacks; and 2) the choice assay, in which fish are given a choice between the model school and a stationary artificial plant. We found low-to-moderate levels of repeatability, ranging from 0.1 to 0.5, in schooling phenotypes. To identify the genomic regions that contribute to differences in schooling tendency, we used QTL mapping in two types of crosses: benthic × marine backcrosses and an F2 intercross. We found two QTL for time spent with the school in the model school assay, and one QTL for number of approaches to the school in the choice assay. These QTL were on three different linkage groups, not previously linked to behavioral differences in sticklebacks. Our results highlight the importance of using multiple crosses and robust behavioral assays to uncover the genetic basis of behavioral variation in natural populations.