Neuropsychological case-studies suggested that dates and encyclopedic numbers may be processed differently than unknown numbers. However, this issue was seldom investigated in healthy participants. Therefore, it is unclear whether known dates are read like words (as lexical items), or like numbers (each position strictly defines digits’ values in a base-10 system). Here, we compared dates to unknown numbers in an experiment using a paradigm from the word recognition literature. We assessed the word-superiority effect by testing experts (students/ teachers in History) with dates. A 4-characters stimulus (xxxx; letters or numbers, half known/unknown) was presented centrally, masked, and followed by 2 characters above and below the mask, at position 2 (xXxx) or 3 (xxXx) in an alternative-forced-choice recognition task. Both accuracy and reaction times were better for dates than unknown numbers, similarly to the results obtained with words by comparison to non-words. However, this effect was modulated by position in the string. These results show a “date-superiority effect” revealing that dates are processed differently than unknown numbers, and suggest that similar orthographical mechanisms might be used to process dates and words.