As the Internet, the global medium of the future, expands exponentially, it has become increasingly relevant to scientific research. So far, there is but little evidence that online testing is suitable for collecting ability-test data. The present article aims to shed light on some aspects of the issue by comparing the performance in a computer-administered ability test of one lab sample and two online samples using a quasi-experimental design. Mean score differences appeared, but can be explained by differences in age and education, and were not due to the test setting (online vs. laboratory). Also, there were no structural differences between the achievement scores of both samples. Some limitations on generalizability are discussed.