Odor processing depends on prior experience, current context and the animal's internal state. To study how learning dynamically changes responses in olfactory cortical circuits, we used an associative learning task in a virtual reality setting. Mice were trained to discriminate one of the four odor-context stimuli to obtain a reward. Previously, we showed that odor-responsive piriform cortex (PCx) neurons become mixed-selective after learning, encoding positional, contextual, and associative information, but how this shift emerges is still unclear. Here, we performed PCx neuronal recordings throughout learning, combining behavioral and neuronal encoding/decoding models to address this question. We show that animals learn the task sequentially, first discriminating odors (intermediate session) and then contexts (expert session), and found that PCx encoding mirrors learning: intermediate session animals only encode odors, while expert animals encode both odors and contexts. Neuronal decoding analyses confirm this result. Moreover, expert animal neurons showed enhanced multiplexing, compared to previous session animals. This evidence indicates that the observed contextual PCx encoding depends on the association of odor cues with positional information to produce correct behavioral responses. Ongoing experiments focus on the neural mechanisms by which this association emerges in the PCx.