In this paper, some commonly used procedures for respondent selection within household, for example, Kish(1949) , Deming(1960) , Troldahl and Carter(1964) and Hong(1996), are modified by using the random selected phone numbers in telephone survey. The modified procedures keep the simplicity and randomness in respondent selection and can be programmed easily under the modern computer assistant telephone interviewing system(CATIS). The goodness-of-fit of our proposed methods are examined by comparing the distributions of the ages and the sex variables of the samples with those of population of Taiwan area. Then suitable selection procedures are chosen using a criterion which measures the goodness of fit.