Cross contamination of intramyocellular lipid (IMCL) signals through loss of bulk magnetic susceptibility (BMS) differences was detected in human muscles using proton magnetic resonance spectroscopic imaging ((1)H-MRSI) at 4 T by varying nominal voxel sizes on healthy subjects. In soleus muscle the IMCL content estimated in 1.00-ml-sized voxels was 15% and 30% higher than that in 0.25-ml voxels for nonobese (P < 0.05) and obese (P < 0.01) subjects, respectively, whereas no effect was observed on IMCL estimation in tibialis posterior (TP) and tibialis anterior (TA) regions for different voxel sizes. The unbiased 0.25-ml voxel size (1)H-MRSI method was applied to measure IMCL content in nonobese sedentary (NOB-Sed), moderately trained (Ath), sedentary obese (OB), and Type 2 diabetic mellitus (DM) subjects. IMCL content in soleus was greatest in OB, with decreasing content in DM, Ath, and NOB-Sed, respectively (12.6 +/- 1.6, 9.7 +/- 1.8, 7.4 +/- 1.0, 4.9 +/- 0.5 mmol/kg wet wt; P < 0.05 by ANOVA; P < 0.05 OB vs. NOB-Sed or Ath). In TA, IMCL was equivalently elevated in DM and OB, which was higher than in Ath or NOB-Sed, respectively (4.2 +/- 0.4 and 4.2 +/- 0.7 vs. 2.7 +/- 0.5 and 1.5 +/- 0.3 mmol/kg wet wt; ANOVA, P < 0.05; P < 0.05 DM or OB vs. NOB-Sed). We conclude that IMCL content is overestimated when voxel size exceeds 0.25 ml despite measurement by optimized high-resolution (1)H-MRSI at high field. When IMCL is measured unbiased by concomitant obesity, we find that it is strongly influenced by muscle type, training status, and the presence of obesity and Type 2 diabetes.