A single flash accompanied by two auditory beeps tends to be perceived as two flashes (Shams et al. Nature 408:788, 2000, Cogn Brain Res 14:147–152, 2002). This phenomenon is known as ‘sound-induced flash illusion.’ Previous neuroimaging studies have shown that this illusion is correlated with modulation of activity in early visual cortical areas (Arden et al. Vision Res 43(23):2469–2478, 2003; Bhattacharya et al. NeuroReport 13:1727–1730, 2002; Shams et al. NeuroReport 12(17):3849–3852, 2001, Neurosci Lett 378(2):76–81, 2005; Watkins et al. Neuroimage 31:1247–1256, 2006, Neuroimage 37:572–578, 2007; Mishra et al. J Neurosci 27(15):4120–4131, 2007). We examined how robust the illusion is by testing whether the frequency of the illusion can be reduced by providing feedback. We found that the sound-induced flash illusion was resistant to feedback training, except when the amount of monetary reward was made dependent on accuracy in performance. However, even in the latter case the participants reported that they still perceived illusory two flashes even though they correctly reported single flash. Moreover, the feedback training effect seemed to disappear once the participants were no longer provided with feedback suggesting a short-lived refinement of discrimination between illusory and physical double flashes rather than vanishing of the illusory percept. These findings indicate that the effect of sound on the perceptual representation of visual stimuli is strong and robust to feedback training, and provide further evidence against decision factors accounting for the sound-induced flash illusion.