`Snakes' are observed in Tore Supra after injection of high velocity solid hydrogen or deuterium pellets ablated inside the q=1 surface. They are detected, immediately after the ablation, as oscillations in the line integrated densities of the central interferometer channels. The corresponding oscillations in the soft X-ray signals detach from the noise about 70 ms later. Snakes survive sawtooth crashes, but are nevertheless affected by them. Variations, during the about 500 ms long lifetime, of the snake radius rs, of the rotation frequency and of the rotation direction are discussed, stressing the effects of the sawtooth crashes. In many snakes rs/r1, where r1 refers to the radius of the q=1 surface, is of the order of 0.5. The snake has an m=1, n=1 helicity, pointing out the problem of the interpretation in terms of current profile. Combined simulations of the snake oscillations in both interferometer and soft X-ray signals have indicated that the impurity (carbon) density inside the snake is much larger than that outside it. Since a change of regime seems to appear about 80 ms after the snake formation on the soft X-ray, it seems plausible that impurity (carbon) accumulation takes place at this time. A stability criterion taking into account both impurity and bootstrap effects is presented; the result agrees with the model proposed by Wesson (1995)