An illusory correlation (IC) is the erroneous perception of a correlation between two actually uncorrelated categories. Empirically, ICs often result from an overestimation of the frequency of the rarest category combination. Robust ICs are found for short learning periods. Yet little is known about whether ICs persist after extended learning periods. Accounts relying on incomplete learning predict that ICs should disappear after extended learning, whereas accounts based on distinctiveness or information loss predict persisting ICs even after prolonged learning periods. The few existing studies using extended learning paradigms produced conflicting results. While some studies using up to 360 trials indicate that the ICs persists even after long learning periods, other studies showed that the IC might already disappear after 90 trials. In this study, the persistence of illusory correlations was tested by exposing participants to a total of 1500 stimuli. One of two visual symbols (frequent or rare symbol; ratio 2:1) was presented concurrently with one of two auditory signals (frequent or rare tone; ratio: 4:1). Across all trials, the symbols were perfectly uncorrelated with the signal. After extended learning, participants overestimated the frequency of the combination rare symbol paired with rare tone relative to the frequency of the combination frequent symbol paired with rare tone indicating the presence of an IC. This shows that IC persist well beyond the trial numbers typically investigated in IC research. In addition, the results are not compatible with incomplete learning accounts, but are consistent with accounts based on distinctiveness or information loss.