Fig. 7. Percentage of species variation explained by three fractions, after controlling for forest type: [a] fallow history only, [b] joint effect of fallow history and rockiness, and [c] and rockiness only. Species are arranged according to their response to the first two fractions. Negative values for the joint fallow-rockiness fraction occur in some cases because this fraction is calculated by subtraction rather than directly estimated. In these cases, the species variation is better explained by rockiness and fallow history together than by the sum of their independent effects (Legendre and Legendre 1998:532-533). Species' codes as in Tables 1 and 2.