Temporal stability of mtDNA haplotype frequencies in American shad stocks: To pool or not to pool across years?

American shad (Alosa sapidissima) were collected in 1992 and 1993 at 19 sites along the North American Atlantic coast and assayed for mtDNA variation (N = 1888). χ² tests resulted in the pooling of two population pairs, yielding 17 potential shad stocks. Sampling across years at a geographic location increased both the number of haplotypes observed and the number of singletons (haplotypes recorded in a single fish) but the proportion of haplotypes that were singletons remained constant, regardless of sample size. χ² tests, analyses of molecular variance, and sample size considerations all indicated that the overall effect of temporal subsampling was negligible and pooling of successive yearly samples from the same location was warranted. When analyzing allocation success, higher informational returns were gained through sampling more fish using fewer (the best six or nine) restriction enzymes than by sampling fewer fish using more enzymes. Allocation success was greater with 17 stocks (temporal subsamples pooled) than with 29 samples (subsamples separate), suggesting that a "pool then allocate" strategy optimizes the accuracy of mixed-stock analysis. Cumulative findings of this and previous American shad studies reinforce the importance of obtaining samples large enough to distinguish between low-frequency polymorphisms, which are informative, and single-occurrence haplotypes, which are not.