High frequency intravascular ultrasound may show a high intensity backscatter from blood which hampers the discrimination between lumen and arterial wall. In this study, the acoustic behaviour of blood at 30 MHz in relation to rouleau size was analyzed. In a Couette viscometer, high frequency (20-40 MHz) backscatter data from normal and modified blood samples from eight volunteers were obtained at shear rates from 0 to 1000 s-1. The acoustic behaviour of blood was quantified by the integrated backscatter power and the spectral slope of the backscatter coefficient. Backscatter from blood depended on rouleau size. At a shear rate of zero, both whole blood and rouleau-enhanced blood showed a 11-dB-higher integrated backscatter power than rouleau-suppressed blood, which itself was 10 dB higher than that of hemolysed blood, the latter showing a 6-dB-higher backscatter than saline. Platelets did not contribute to the backscatter power. Plasma and saline produced no detectable integrated backscatter power other than noise. The spectral slope of whole and rouleau-enhanced blood was small (1 and 0.5, respectively), whereas rouleau-suppressed blood and hemolysed blood (both with a slope of 3.3) behaved almost like a Rayleigh scattering medium (slope = 4). The backscatter from rouleau-suppressed blood showed no shear rate dependence. At low shear rates ( < 0.8 s-1 for integrated backscatter power and < 0.2 s-1 for the spectral slope), whole blood and rouleau-enhanced blood tended to the results from the static situation (no shear). At high shear rates ( > 80 s-1 for integrated backscatter power and >11 s-1 for spectral slope), these samples tended to the results of rouleau-suppressed blood. Ultrasound backscatter at 30 MHz from human blood is only caused by red blood cells. With increasing aggregate (rouleau) size, the integrated backscatter power increased by 11 dB, and the spectral slope decreased from 3.3 to 1.