The Ten-Times Rule: A Gap Made of Sampling

DNA barcoding promised a master key: read 648 base pairs of the mitochondrial COI gene and a specimen names its species. The promise of discovery by distance rested on the barcoding gap, and from it came Hebert and co-authors' ten-times rule (birds, PLoS Biology 2004): flag a lineage as a provisional species when its divergence exceeds ten times the group's mean intraspecific variation, a line at 2.7 percent on the bird base of 0.27 percent, against a congeneric mean of 7.93 percent. This paper censuses the rule against the first comprehensively sampled test, Meyer and Paulay's 2005 cowrie study: 2,026 individuals, 263 evolutionary units, more than 93 percent of a family's recognized species. Identification of unknowns against a complete reference phylogeny errs as low as 4 percent; threshold-based delineation errs at a minimum of about 17 percent in cowries, at about 34 percent under the ten-times rule itself, and at 33 percent when only traditionally recognized species are analyzed. The printed mechanism is sampling from both sides: one to two individuals per species understated variation within species, and regional sampling missed the allopatric sister taxa that fill the gap from above. Grade: **THE GAP WAS REAL IN THE SAMPLE AND AN ARTIFACT OF IT — PRICED AT SEVENTEEN PERCENT WHERE THE SAMPLING WAS HONEST.** This paper belongs to the breadth program; it touches no claim of the physics core.