We show that in the derivation of the frequency of Thomas precession, the fact of implementation of rotation-free Lorentz transformations between a laboratory frame, KL, and Lorentz frames K(t) co-moving with a particle with spin at any time moments, t, has principal importance. Choosing for the observation of the particle’s motion any other inertial frame, K, related with KL by the rotation-free transformation, we have to realize that the transformations between K and K(t) at any t are no longer rotation-free. This way we provide a resolution of the known paradox by Bacry (H. Bacry. Nuovo Cimento, 26, 1164 (1962)) and suggest a reinterpretation of the Thomas precession, which is further discussed.