The nature of vascular contribution to blood oxygenation level dependent (BOLD) contrast used in functional MRI (fMRI) is poorly understood. To investigate vascular contributions at an ultrahigh magnetic field of 9.4 T, diffusion-weighted fMRI techniques were used in a rat forepaw stimulation model. Tissue and blood T(2) values were measured to optimize the echo time for fMRI. The T(2) of arterial blood was 40.8 +/- 3.4 msec (mean +/- SD; n = 5), similar to the tissue T(2) of 38.6 +/- 2.1 msec (n = 16). In comparison, the T(2) of venous blood at an oxygenation level of 79.6 +/- 6.1% was 9. 2 +/- 2.3 msec (n = 11). The optimal spin-echo time of 40 msec was confirmed from echo-time dependency fMRI studies. The intravascular contribution was examined using a graded diffusion-weighted spin-echo echo-planar imaging technique with diffusion weighting factor (b) values of up to 1200 sec/mm(2). Relative BOLD signal changes induced by forepaw stimulation showed no dependence on the strength or direction of the diffusion-sensitizing gradients, suggesting that the large vessel contribution to the BOLD signal is negligible at 9.4 T. However, gradient-echo fMRI performed with bipolar diffusion sensitizing gradients, which suppress intravascular components from large vessels, showed higher percent signal changes in the surface of the brain. This effect was attributed to the extravascular contribution from large vessels. These findings demonstrate that caution should be exercised when interpreting that higher percent changes obtained with gradient-echo BOLD fMRI are related to stronger neural activation. Magn Reson Med 42:919-928, 1999.