This study assesses the changes in genetic diversity using two different methods in eight captive bred lines of banana shrimp (Fenneropenaeus merguiensis) that had been mass selected for length for up to 14 generations. Specifically, mitochondrial D-loop DNA sequencing and genotyping using five DNA microsatellite loci were used to document changes in haplotype diversity and allelic diversity numbers at several time points during and up to 14 generations of captive bred lines, typically maintained without intercrossing among lines. Data from eight of the lines were compared with each other and to a reference sample of wild caught animals. As each wild animal had a unique mtDNA haplotype, we estimate there were 20 different mtDNA haplotypes in each of the two different founding stocks. The average number of haplotypes was 1.8 after 11 or more generations of captive breeding. Similarly, whereas the wild reference stock had an average number of more than 13 microsatellite alleles per locus, the descendent lines had an average of 5.6 per locus after 11 or more generations. These declines were evident despite strategies that had been put in place to maintain genetic variation, including the use of up to 1000 brood stock per generation. The loss of genetic variation was unequivocal being evident for both DNA methods and in all the different lines. Effective population size (Ne), as derived from linkage disequilibrium, was estimated to average about nine after 11-13 generations in the captively bred lines, compared with 263 estimated in the wild samples. This corresponds to a rate of inbreeding of about 4% per generation for the captively bred lines. Additive genetic variance of the captively bred lines, estimated under the assumption of neutrality, ranged from about 75% to 25% that in the wild samples. We therefore conclude that mass selection, even when using a relatively large number of broodstock, still results in substantial loss of allelic diversity within lines over generations, and a reduction of effective population size and genetic variance, to the degree that productivity could have been compromised compared with similarly selected but outbred stocks. Loss occurred relatively consistently among the different lines. It was common for different microsatellite alleles or mtDNA haplotypes to have persisted in the different lines, such that the total number of haplotypes and allele types among all lines was much greater than that within given single lines, and the number of alleles among lines approximated that found in the wild. This observation, evident because many different lines were monitored, suggests that under certain circumstances (fixation and selection), more net genetic variability can be maintained over many generations of selection by keeping multiple different and independent lines rather than one large single line. Accordingly, if multiple lines are maintained, there could be some practical options to reconstitute allelic and haplotype variation without new introductions of genetically unimproved stock from the wild.