The purpose of the research presented here was to study the temporal transferability of mode choice models for work trips and other home-based trips based on the mobility surveys conducted in the Helsinki Metropolitan Area in 1988 and 1995. The updating procedures examined were the Bayesian updating, combined transfer estimation, transfer scaling, and joint context estimation procedures. To explore the impact of sample size on transferring performance, model transferability was tested using three to four different sample sizes. Hence all the procedures were examined using 100 samples for each trip group and sample size category. The model transferability was examined by comparing the transferred models to the models estimated using the entire set of the data which can be regarded as the best estimate representing "the real situation". The results show that the Helsinki Metropolitan Area models are transferable. The best transferred model produced, in all cases, better results than the corresponding model based on a small sample. However, finding the correct method and sample size for each case is not an unambiguous task. The appropriate method and sample size depend both on the difference between model coefficients in the initial and the final stages and the quality of the data. The biggest problem in model transferring is the great confidence interval of the unstable parameters. The more the new data are emphasised, the greater is the variation of model coefficients and the risk of obtaining wrong coefficients. This is also the reason, why the model transferability was tested as variable-oriented based on the accuracy of different coefficients. Generally, joint context estimation was the best method. In particular, the method was useful, if the transfer bias was large or only some of the coefficients were accurate. The transfer scaling also gave rather good results. The problem with this method was the unpredictability of the quality of results. The scale factor is usually estimated for travel time and the number of transfers. If the ratio of these coefficients does not stay, the model's ability to predict the effects of changes in transportation system may become rather weak. The combined transfer estimation procedure performance was best when there was a large number of observations and the transfer bias was small. The Bayesian approach emphasises the coefficients with respect to the inverse of the variances of each coefficient. Therefore, the coefficients based on the estimation context data may be emphasised too strongly in model transferring, since the number of observations is larger in this context.