The autocorrelation parameter of AR(1) model is estimated very often by the ordinary least squares estimator (OLSE) due to its simplicity. The present investigation aims at deriving the algebraic expression of the covariance between two OLSE's obtainable from two overlapping (OS) or non-overlapping (NOS) or gapping (GS) series whatsoever choosing from the given whole series. Such expression is used to obtain the expressions of bias, mean square error and variance of Quenouille's estimator (1956). Based on OS splitting, a Quenouille-type family and another competent family of estimators are suggested. Their comparative performances are discussed in respect of bias and mean square error.