Tournament Structure
Chameleons were initially grouped into size-matched
quads based on SVL and mass (mean % difference
between members within quads SE ¼ 2.6 0.4% forSVL and 10.4 1.3% for mass). Within each quad, each
member contested every other member such that each individual
was initially in three contests (six contests per
quad: 1 versus 2; 1 versus 3; 1 versus 4; 2 versus 3; 2 versus
4; 3 versus 4). We then conducted contests between members
of different quads: members of smaller size class
quads were matched against members from the next larger
or smaller size class quad, depending on whether they had
won or lost within-quad contests. If a winner from
a smaller quad beat the winner from a larger quad, he
was pitched against an individual from the next larger
quad, and the converse for losers. We planned our contests
according to a ‘tournament structure’ whereby each
individual contested both similar-sized and different-sized
individuals, based on our knowledge of previous wins and
losses for each individual. This design ensured a robust
data set for application of the B–T model (see below). To
avoid stressing the animals, a minimum of 2 days, usually
more (X SE¼ 6:68 0:63 days) was allowed between
consecutive trials.