A robust finding in the literature is that spacing material leads to better retention than massing; however, the benefit of spacing for concept learning is less clear. When items are massed, it may help the learner to discover the relationship between instances, leading to better abstraction of the underlying concept. Two experiments addressed this question through a typical function learning task in which subjects were trained via presentations of input points (cue values) for which output responses (criterion values) were required. Subjects were trained either using spaced points, strategically massed points (points were paired in training such that they occurred on the same side of the underlying V-shaped function), or randomly massed points (points were randomly paired during training). All subjects were then tested on repeated training points, new (interpolation) points within the training range, and extrapolation points that fell outside the training range. Spacing led to superior interpolation and extrapolation performance, with random massing leading to the worst performance on all test trial types. These results suggest that, at least for function concepts, massed training is not superior to spaced training for concept learning.