Fine-needle aspiration (FNA) remains the gold standard for the diagnosis of thyroid cancer. However, currently, a large number of FNA biopsies result in negative or undetermined diagnosis, which suggests that better noninvasive tools are needed for the clinical management of thyroid cancer. Spectral-based quantitative ultrasound (QUS) characterizations may offer a better diagnostic management as previously demonstrated in mouse cancer models ex vivo. As a first step toward understanding the potential of QUS markers for thyroid disease management, this paper deals with the spectral-based QUS estimation of healthy human thyroids in vivo. Twenty volunteers were inspected by a trained radiologist using two ultrasonic imaging systems, which allowed them to acquire radio-frequency data spanning the 3-16-MHz frequency range. Estimates of attenuation coefficient slope (ACS) using the spectral logarithmic difference method had an average value of 1.69 dB(cṁMHz) with a standard deviation of 0.28 dB(cṁMHz). Estimates of backscatter coefficient (BSC) using the reference-phantom method had an average value of 0.18sr-1cm-1 over the useful frequency range. The intersubject variability when estimating BSCs was less than 1.5 dB over the analysis frequency range. Further, the effectiveness of three scattering models (i.e., fluid sphere, Gaussian, and exponential form factors) when fitting the experimentally estimated BSCs was assessed. The exponential form factor was found to provide the best overall goodness of fit (R2=0.917), followed by the Gaussian (R2=0.807) and the fluid-sphere models (R2=0.752). For all scattering models used in this study, average estimates of the effective scatterer diameter were between 44 and 56μm. Overall, an excellent agreement in the estimated attenuation and BSCs with both scanners was exhibited. © 2016 IEEE.