This paper presents two techniques in iterative learning identification (ILI) when the zero initial state condition is not achieved. One is to obtain acceptable impulse responses. The other is to measure the response error to the exclusion of non-zero initial state factor. This paper proposes an estimation technique using the least-squares (LS) method for the former and introduces discarded data in measurement of the response error for the latter. The ILI with the proposed techniques is applied to estimation of the aerodynamic derivatives in a lateral linear model of aircraft. The effectiveness of the proposed techniques is demonstrated in numerical simulations. © 2014 The Japan Society for Aeronautical and Space Sciences.