Abstract In two experiments, we used the familiarization-novelty preference procedure to investigate the mechanism producing the pattern-line effect previously observed in 3- and 4-month-old infants. In the first experiment, we demonstrated the pattern-line effect as evidenced by the significantly higher novelty preference when infants were both familiarized and tested with mirror-image line elements embedded in a right-angle framework than when the infants were familiarized and tested with the same-line elements in isolation. We also observed a significant novelty preference when infants were familiarized with an isolated-line element, but tested with the elements embedded in context and a marginally significant preference when infants were familiarized with an element in context and then tested with the isolated elements. These results cast doubt on a strict emergent feature explanation of the pattern-line effect. In a second experiment, infants were shown to generalize habituation on the basis of configural information. The overall pattern of results is consistent with both the global-to-local ( Navon, 1977) and levels of information ( Enns & Prinzmetal, 1984) accounts of the pattern-line effect.