There is an ongoing debate on how the meaning of words is retrieved in a newly-learned language L2. Two processes are assumed: L2 words retrieve underlying concepts via associations with the corresponding words in the first language, L1, (word association hypothesis) or L2 words gain direct access to underlying concepts (concept mediation hypothesis). It is commonly shown that a higher language proficiency is needed to access word meanings directy via concepts. However, evidence for newly-learned words and the initial learning of novices are scarce. Using an item-specific priming paradigm, we demonstrate that classes and actions (e.g., “small, right!”) merely instructed during an item’s prime lead to repetition priming effects (i. e., reduced reaction times for item-specific class/action mapping repetitions) in the item’s subsequent probe (lag 2-7 trials). Crucially, this was the case to a comparable degree both when participants were instructed in L1 and when they were instructed in an L2 which they had no knowledge of prior to a brief practice phase at the beginning of the experiment. These findings indicate that the direct route to concepts can be accessed already in the very beginning of language learning and highlight how fast newly-learned words become able to automatize behavior.