The aim of this study was to investigate the use of (1)H-MRS with various echo times to characterize subcutaneous human adipose tissue (SAT) triglyceride composition and to validate the findings with fatty acid (FA) analysis of SAT biopsies by gas chromatography (GC). (1)H-MRS spectra were acquired with a 1.5 Tesla clinical imager from the SAT of 17 healthy volunteers, with 10 undergoing SAT biopsy. Spectra were localized with PRESS and a series of echo times; 30, 50, 80, 135, 200, 300 and 540 ms were acquired with TR = 3000 ms. Prior knowledge from phantom measurements was used to construct AMARES fitting models for the lipid spectra. SAT FA composition were compared with serum lipid levels and subject characteristics in 17 subjects.Long TE (135, 200 ms) spectra corresponded better with the GC data than short TE (30, 50 ms) spectra. TE = 135 ms was found optimal for determining diallylic content (R = 0.952, p < 0.001) and TE = 200 ms was optimal for determining olefinic content (R = 0.800, p < 0.01). The improved performance of long TE spectra is a result of an improved baseline and better peak separation, due to J-modulation and suppression of water. The peak position of the diallylic resonance correlated with the average double bond content of polyunsatured fatty acids with R = 0.899 (p < 0.005). The apparent T(2) of the methylene resonance displayed relatively small inter-individual variation, 88.1 +/- 1.1 ms (mean +/- SD). The outer methyl triplet line of omega-3 PUFA at 1.08 ppm could be readily detected and quantitated from spectra obtained at TE = 540. The omega-3 resonance correlated with the omega-3 content determined by GC with R = 0.737 (p < 0.05, n = 8). Age correlated significantly with SAT diallylic content (R = 0.569, p = 0.017, n = 17), but serum lipid levels showed no apparent relation to SAT FA composition. We conclude that long TE (1)H-MRS provides a robust non-invasive method for characterizing adipose tissue triglycerides in vivo.