An identification system for individual snakes, Natrix maura, was developed, based on photographs of the pattern on the ventral scales. The patterns appeared to be constant, and contained sufficient information (allowing for non-randomness within the pattern) to render duplication unlikely, even in extremely large populations. After digitization, the patterns were compared for overall similarity by computer, and those which matched were checked visually. Using this system on a third of the scales, only 1% of known recaptures (those in which one of the compared photographs was of poor quality) were missed, and coding was reduced to 2.2 min. This sifting makes recognition by natural marks practicable in very large samples, and is robust enough to absorb considerable error in coding.