Rolling amplitude of a ship decreases as the natural period increase. This fact is true not only on smooth swells, but also on confused seas. If we suppose that there are two ships simultaneously whose each period is T ₁ and T ₂, then the ratio of expected maximum value of rolling amplitude is approximately θ₁/θ₂=1/( T₁ / T₂ )² so that if T₁2, θ₁ is considerablly larger than θ₂. The author examined these results on the long records of weather ships at X-ray (39 N, 153 E) and Tango (29 N, 135 E), whose types are different and the period is about 7.60 sec. and 12.32 sec respectively. For a ship whose period is long, rolling amplitudes are comparatively stationally and regular as the ship tends to prefer and select swells, but on the contrary to this, for a ship which has short period resonance waves are usally confused seas, so that statistical amplitudes are not narrow banded but widely spread and these distributions are the same as that of Drs. Longuet-Higgins.