Increasing demand for higher precision machines is arising in the high-tech industry. Linear guides are often used to establish this highly accurate and highly repeatable motion. The amount and variation in preload play an important role in the total motion errors. The objective of this paper is to investigate the relation between preload and repeatability for both the ball linear guide and the cross-roller linear guide by performing experimental tests. Three preload cases at three positions are investigated and both longitudinal repeatability and rotational repeatability are measured. In the first case, the preload will be applied uniformly by adjusting two adjustment screws on one side. In the second case, only one screw will be adjusted and therefore generates an angular preload. In the third case, an additional mass is added resulting in external preload. The experiments are performed at 0-25mm, 25-50mm, and 50-75mm, in which zero represents the position both rails are in one line. Performances of the roller guide have a smaller spread at different positions compared to the ball guide. Also, after averaging the different positions for both rails, the repeatability of the roller guide tends to be generally lower than the ball guide, suggesting that the performances of the roller guide are better than those of the ball guide. For the ball guide, the longitudinal repeatability has an optimal point for uniform applied preload, while for the roller guide the performance slightly improves as the preload increases. The longitudinal repeatability generally increases for both the ball and roller guide with the angular applied preload, with the exception of an adjustment of 10 degrees for the ball guide. The longitudinal repeatability stays constant with the external preload. The rotational repeatability decreases for both the ball and roller guide with both uniform and angular applied preload. The rotational repeatability remains constant for both the ball and roller guide with external preload.