This paper reports an investigation of the nature of allophonic variation in the single liquid phoneme of standard Korean. Alveolar tap and alveolar lateral allophones are in strict complementation, and an intervocalic length contrast of singleton tap vs. geminate lateral also arises on the surface. These are sometimes cited as reasons Koreans do better at the English /l/-/r/ distinction than other learners who similarly lack an underlying L1 contrast. To investigate native perception of the two intervocalic possibilities, waveform editing was used to eliminate the duration difference between geminate laterals and singleton taps in recordings of natural speech. In a forced-choice test, all listeners identify the edited stimuli as containing the lateral (90.5% identification rate), suggesting that duration is not a deciding factor in identification. Instead, Korean L1 speakers appear sensitive to non-durational differences, and thus effectively have a latent /l/-/r/ contrast.