Abstract Operator response times for off-normal events in nuclear power plants have been used to assess human failure probabilities. Usually, log-normal distribution is applied to the relation of response times and failure probabilities. In the literature there have been several studies on estimating the response times. Since the response times are affected by the performance shaping factors, a regression model can be applied to estimate the response times. The conventional regression model regards the deviations between the observed and estimated values as measurement errors. However, the deviations of the dependent variable, the response times can be dependent upon the fuzziness of the parameters for the independent variables, the performance shaping factors. In this research, possibilistic linear regression models have been used to asses fuzzy parameters and furthermore response times associated with the performance shaping factors.