Recent studies demonstrate perceptual learning of unattended stimuli that precede or predict the explicit task target. Implicit learning has improved visual coherent motion detection thresholds [1–3] and auditory categorization of artificial structured sound groups [4]. Based on these findings, we attempted to train Greek listeners to distinguish a Hindi dental-retroflex stop consonant contrast without task awareness or feedback, extending previous findings [5] challenging the standard assumption that non-native speech categories are learned by adults only when trained in focused attention with explicit feedback. Perceptual pre-tests indicated that Greeks are initially unable to differentiate the two Hindi categories but can learn the contrast with standard training. Subjects engaged in a demanding voice identification task in ten daily sessions. They heard rapid Greek syllable sequences in a variety of voices and had to repeat the syllable pronounced by one of two target voices. Hindi syllables with retroflex sounds, spoken by a single native speaker, always preceded the target voices, while Hindi dentals appeared in non-predictive positions. After training, subjects were tested in dental-retroflex discrimination and identification, and were compared to untrained Greek listeners. To test whether learning is a result of mere exposure to the Hindi sounds, control groups are tested with the same materials except for the co-occurrence of Hindi retroflex with the task target. References [1] Watanabe T et al., Nature, 413: 844–848. 2001. [2] Watanabe T et al, Nature Neurosci, 5, 1003–1009, 2002. [3] Seitz AR & Watanabe T, Nature, 422, 36, 2003. [4] Wade T & Holt LL, J Acoust Soc Am, 118, 2618–2633, 2005. [5] McCandliss BD et al., Cogn Affect Behav Neurosci, 2, 89–108, 2002.