We assessed language switch and mixing costs in a language-general semantic categorization task and examined how these costs relate to general inhibition and set shifting capacities. The participants were 51 native Finnish subjects with English as L2. The results showed significant symmetric language switch costs and, unexpectedly, a mixing advantage in L2: reaction times were faster in the mixed language block than in the single language block. The interactions with the general executive functions showed no consistent overall pattern. We argue that the L2 mixing advantage stems from statistical facilitation in line with a horse race model, or from opportunistic planning as suggested by the Adaptive Control hypothesis. We argue that the results overall indicate that lexical access in language reception is non-selective.