Identifying and reporting genetic variability is an important tool to increase the utilization of germplasm collections.  In an effort to further realize the potential of cottonseed, we have characterized seed composition traits of Gossypium spp. accessions available in the U.S. National Cotton Germplasm Collection (NCGC).  This characterization will indicate the diversity that can be found in these resources and allow researchers to further explore nutritional and other value-added uses for this agricultural commodity.  Seed quality analyses were conducted to explore relationships between oil and protein reserves using low-field, time-domain ¹H nuclear magnetic resonance (TD-NMR).  This non-destructive method was used to quantify oil and protein content within 2,254 accessions of cotton including wild species, landraces, and adapted cultivars from around the world as represented in the NCGC.  This group of accessions is collectively known as the Gossypium Diversity Reference Set (GDRS) and represents ~20% of the content of the NCGC, comprising a range of genomes (one tetraploid and eight diploid) and species (five tetraploid and 28 diploid).  Our analyses show a wide range of oil (22.8% (TX-0460) to 8.77% (TX-1688)) and protein (33.4% (SA-1254) to 14.2% (TX-2350)) content.  Genome K had the highest mean oil percent (21.4%) while the cultivated diploid A genome had the overall highest percentage of protein (24.3%).  The AD genome (including the commercially cultivated tetraploid species, G. barbadense and G. hirsutum) had the third highest oil content (18.7%) and the second highest protein content (23.3%).  In general, genomes C, E, and F ranked low for these seed quality measurements.  Genome B had a noticeably low mean for oil content (12.8%) while genome C was noticeably the lowest in terms of protein content (14.4%).  The correlation between oil and protein for this entire data set indicated there was an inverse relationship between these two traits (r = -0.29).  However when we look closer at correlations within genomes and species, we see a large amount of variability in this relationship.  For the species currently cultivated on a commercial-scale, the correlation was negative for G. arboreum (r = -0.15), G. barbadense (r = -0.25), and G. hirsutum (r = -0.55), with no correlation for G. herbaceum (r = 0.02).  However, within these species, individual accessions can be found for every combination of oil and protein content; thereby indicating the variability present that can be manipulated to enhance these traits.  These analyses will provide a more comprehensive knowledge of cottonseed components to facilitate effective identification and utilization of accessions by breeding programs interested in augmenting these seed traits.