Note that A actually represents the age at tagging, a1, relative toa0(i.e., A = a1− a0), since a0cannot be separated from a1using thetagging data alone. A is assumed to follow a specified distribution,and the parameters of this distribution are estimated within themodel. In applying the LEP method to the three tuna data sets, alognormal distribution with parameters logAand logAwas cho-sen for A. Laslett et al. (2002) showed that the results were fairlyrobust to the distribution chosen for A so long as it provided areasonable approximation. Another feature of the LEP method isthat it allows for individual variability in growth by modelling theasymptotic length parameter as a random effect. For all species, L∞was assumed to follow a normal distribution with mean L∞andstandard deviation L∞. The model allows for measurement errorand/or additional process error in length at age through an addi-tive Gaussian error component with mean 0 and standard deviationtag.Conditional on a known value of A, say A = a, the release length(L1) and recapture length (L2) are both the sum of Gaussian nor-mal variables (one component relating to L∞and one relating tomeasurement error), and are therefore Gaussian themselves withmean, variance and covariance: