The human face is complex and multipartite, and characterization of its genetic architecture remains intriguingly challenging. Applying GWAS to multivariate shape 2 phenotypes, we identified 203 genomic regions associated with normal-range facial variation, 117 of which are novel. The associated regions are enriched for both genes relevant to 4 craniofacial and limb morphogenesis and enhancer activity in cranial neural crest cells and craniofacial tissues. Genetic variants grouped by their contribution to similar aspects of facial 6 variation show high within-group correlation of enhancer activity, and four SNP pairs display evidence of epistasis, indicating potentially coordinated actions of variants within the same cell 8 types or tissues. In sum, our analyses provide new insights for understanding how complex morphological traits are shaped by both individual and coordinated genetic actions. 10 Main Text: 12 "One of the major problems confronting modern biology is to understand how complex morphological structures arise during development and how they are altered during evolution""complicated developmental choreography" in which intrinsic genetic factors, epigenetic factors, and interactions between the two make up the progeny genotype, which engages with the 20 environment to ultimately produce a complex morphological trait, defined thus by its composition from a number of separate component parts 1 . We now understand that the intrinsic 22 genetic factors ultimately contributing to complex morphological traits consist not only of single 2 variants altering protein structure and/or function, but also non-coding variants and interactions 24 among variants, each affecting multiple tissues and developmental timepoints. This realization necessitates the development and utilization of methods capable of describing the genetic 26 architecture of complex morphological traits, which includes identifying the individual genetic variants contributing to morphological variation as well as their interactions 2,3 . 28 The human face is an exemplar complex morphological structure. It is a highly multipartite structure resulting from the intricate coordination of genetic, cellular, and 30 environmental factors 4-6 . Through prior genetic association studies of quantitative traits, 51 loci have been implicated in normal-range craniofacial morphology, and an additional 50 loci have 32 been associated with self-reported nose size or chin dimples in a large cohort study 7 (Table S1).However, as with all complex morphological traits, our ability to identify and describe the 34 genetic architecture of the face is limited by our ability to accurately characterize its phenotypic variation 4 , identify variants of both large and small effect 8 , and identify interactions between 36 variants. We previously described a novel data-driven approach to facial phenotyping, which facilitated the identification and replication of 15 loci involved in global-to-local variation in 38 facial morphology 9 . Here, we apply this phenotyping approach...