The problem of testing for symmetry of a continuous distribution about a specified median is considered. The runs test proposed by McWilliams (McWilliams, T. P. (1990). A distribution-free test for symmetry based on a runs statistic. Journal of the American Statistical Association, 85(412), 1130-1133) is modified by conditioning on the number of observations greater than the specified median. Simulations show that the modified test slightly outperforms the unconditional one in terms of power and is quite robust for possible misspecification of the median.