Summary: Kernel smoothing is studied in partial linear models, i.e. semiparametric models of the form \[ y_ i=\xi_ i'\beta +f(t_ i)+\epsilon_ i\quad (1\leq i\leq n), \] where the \(\xi_ i\) are fixed known p vectors, \(\beta\) is an unknown vector parameter and f is a smooth but unknown function. Two methods of estimating \(\beta\) and f are considered, one related to partial smoothing splines and the other motivated by partial residual analysis. Under suitable assumptions, the asymptotic bias and variance are obtained for both methods, and it is shown that estimating \(\beta\) by partial residuals results in improved bias with no asymptotic loss in variance. Application to analysis of covariance is made, and several examples are presented.