Quantitative ultrasound (QUS) is a simple, inexpensive and radiation-free technique to study skeletal status in osteoporosis. Broadband ultrasound attenuation (BUA) is typically used without normalizing for bone thickness. We investigated the impact of bone thickness on BUA, both in in vitro and in vivo, using commercially available ultrasound scanners (Walker Sonix UBA 575 and 575+). For the study in vitro, we analyzed 12 paired cubes (12 x 12 x 12 mm3) of bovine trabecular bone cut from six "parent" blocks (24 x 12 x 12 mm3). These "parent" blocks showed a fairly homogeneous structure as assessed by conventional radiography. BUA and dual x-ray absorptiometry (DXA) measurements of bone mineral density (BMD) were obtained. The difference between the paired cubes was significant for BUA (p < 0.001), but not for BMD (p > 0.05). The BUA of the "parent" blocks was larger by 32-92% than the average BUA of the corresponding paired cubes, and always less than the sum of the paired cubes (5-34%). BUA thus added up in a complex and nonlinear fashion. Measurements in vivo of the calcaneal bone width demonstrated on radiographs (study in vivo I, 28 postmenopausal women) and CT images (study in vivo II, 17 women) showed a small, positive but not significant correlation with BUA (r2 = 0.13 and r2 = 0.007, p > 0.05). The magnitude of this effect provides further evidence that BUA does not scale linearly with bone size. Our results indicate that simple normalization of BUA data by specifying the results in decibels per megahertz per millimeter units may not be appropriate.(ABSTRACT TRUNCATED AT 250 WORDS)