The results of reduced-scale failure tests of simply supported four-point-bend beams of different sizes, containing lapped bond splices of smooth (undeformed) longitudinal reinforcing bars, are reported. The tests consist of two groups, with splices located either in the midspan region with a uniform bending moment, or in one of the end regions with a uniform shear force. The specimens were made of microconcrete with a maximum aggregate size 4.76 mm. Beams of three different heights (50, 100, and 200 mm) were tested. The beams were geometrically similar in three dimensions, and even the bar diameters and cover thicknesses were scaled in proportion. The reinforcement ratio was 0.31%. The results reveal the existence of a significant size effect, which can be approximately described by the size effect law previously proposed by Bazant. The size effect is found to be stronger for splices without any spiral than for splices confined by a spiral, and stronger for splices in the maximum shear force region of a beam than for splices in the maximum bending moment region. Generalization of the existing formula of Orangun et al. is proposed and recommended for design. Although the formula provides a safer alternative to the existing approach, further testing is needed for better calibration. The size effect on the nominal bond strength implied by the development length provisions of the current and previous American Concrete institute code :is discussed and shown to be inadequate.