In the event of two-phase sampling, the authors have constructed a regression-type estimator of population mean of the main variable (y) in the presence of available knowledge on second auxiliary variable (z), when the population mean of the first auxiliary variable (x) is not known. The proposed estimator is found to be more efficient than the usual two-phase ratio and regression estimators and also two-phase ratio-type and regression-type estimators using two auxiliary variables suggested by Mohanty (1967), Chand (1975), Kiregyera (1980, 1984) and Shaoo et al. (1993)