Assessing unconscious processing requires a valid measure of conscious perception. However, the two measures most commonly used, subjective reports and forced-choice discrimination, do not always converge: observers can discriminate stimuli rated as invisible better than chance. A debated issue is whether this phenomenon indicates that subjective reports of unawareness are contaminated by conscious perception, or that forced-choice discrimination performance is contaminated by unconscious processing. To address this question, we took advantage of a previously reported dissociation using masked response priming: for primes rated as invisible on a multi-point scale, response priming occurs only for fast trials, whereas for consciously perceived primes, response priming occurs across response times. Here, we replicated this dissociation, confirming that invisibility-reports were not contaminated by conscious perception. Crucially, we measured prime-discrimination performance within the same experiment and found above-chance performance for unseen primes. Together, these findings suggest that forced-choice discrimination performance is contaminated by unconscious processing.