A classic example of natural selection, that of color-pattern variatio
n in Lake Erie island populations of water snakes, was reexamined to o
vercome shortcomings resulting from classification of snakes into disc
rete color-pattern categories and use of cross-sectional data. Four co
ntinuously varying color-pattern components (DB, the number of dorsal
blotches; LB, the number of lateral blotches; ROWS, the height of late
ral blotches measured in scale rows; and VEXT, the extent of ventral p
igmentation) were analyzed. Patterns of natural selection were predict
ed from the relationship between color-pattern scores and independent
measures of relative crypsis. Tests for natural selection were carried
out using longitudinal data on neonate to juvenile-aged snakes and cr
oss-sectional data on juvenile to adult-aged snakes. As predicted, the
form of selection differed between younger and older age classes of s
nakes: selection resulted in a reduction in DB and LB among neonate an
d juvenile snakes but had little influence on color-pattern components
in older snakes. The correspondence between observed patterns of natu
ral selection and predictions based on the relationship between color-
pattern scores and relative crypsis supports the hypothesis that diffe
rential predation by visual predators on younger age classes of snakes
is the mechanism of selection. Gene flow from mainland populations or
the initial lack of an allele necessary for reduced pattern may expla
in why selection has not resulted in greater differentiation between i
sland and mainland populations.