Radio-control, free-sailing model experiment provides many useful informations on ship design concerning manoeuvrability. In the analysis of such experiments, a proper mathematical model of ship response should be adopted to interprete the test result. Then we can make up a general description of manoeuvrability of that design, not merely giving a direct record of particular manoeuvres actually performed. The simplest of all is the KT -analysis on zig-zag tests. The mathematical model is T ψ+ψ= K δ (1) where ψ denotes yaw-rate and δ rudder angle. This KT -model is, however, too simple to include any non-linearity in ship behaviour. It can not take account of higher-order time constants either, which are often significant in the synthesis of automatic steering control. We have previously presented a mathematical model of steering response of a ship, i. e, T ₁ T ₂ψ+ ( T ₁+ T ₂)ψ+ψ+αψ³= K δ+ KT ₃ψ (2) This paper relates to a procedure of determining the response parameters K , T ₁, T ₂, T ₃, and α from free-sailing model experiments. That is : (1) the yaw gain K is best determined by the reverse spiral tests only at small range of yaw-rate and (2) the time constants T ₁, T ₂, and T ₃ as well as the nonlinear parameter α should then be determined through a least square error iteration analysis on a zig-zag steering test result. Such analysis on four models and eight actual ships are presented.