Population and community descriptors that might be used as indicators of the impact of fishing are reviewed. The criteria used for the evaluation of these indicators are meaning, expected effect of fishing, exclusiveness to fishing effects, and measurability. Population indicators such as total mortality rate, exploitation rate, or average length are the most operational indicators because their meaning is clear and the expected effect of fishing on them is well understood so that reference points can be set. On the other hand, indicators based on the composition of species assemblages such as diversity indices and ordination of species abundances are difficult to interpret, and the effect of fishing on them is not easily predicted. Robust indicators describing the community functions of interest (production and transfer of biomass to large fish), such as size spectra descriptors or the proportion of piscivorous fish in the community, are more promising but are not yet well developed. New candidate indicators are proposed: the change in fishing mortality required to reverse population growth rate, the proportion of noncommercial species in the community, and the average length and weight in the community.

Feeding preferences of Celtic Sea fishes were investigated using a database of stomach content records, collected between 1977 and 1994. The diet of cod Gadus morhua, hake Merluccius merluccius, megrim Lepidorhombus whiffiagonis, whiting Merlangius merlangus and saithe Pollachius virens changed markedly as the animals grew larger, and although large predators generally chose larger bodied prey, the variability of prey sizes consumed also increased. Large predators continued to select small, low value, benthic prey (e.g. Callionymus spp. and Trisopterus spp.) which were easier to catch, rather than larger, more energy lucrative pelagic prey (e.g. mackerel Scomber scombrus), even though these pelagic prey-fishes were nearly always available and were often very abundant. Stock estimates of the International Council for the Exploration of the Sea and U.K. groundfish survey catches were used as indices of prey abundance. Blue-whiting Micromesistius poutassou and other small pelagic fishes (Argentina spp. and clupeoids) were identified as being particularly important, and were consumed by some predators more often than would be expected given the abundance of these prey in the environment. There was no evidence for density-dependent feeding by predators on mackerel and only hake exhibited densitydependent feeding on horse-mackerel. Hake, cod and megrim consumed more blue-whiting when this prey was at higher abundance in the environment. In choosing what prey to consume, predators must balance costs and benefits, considering the quality of prey and the energy expended during search, capture and handling.

Sustainable exploitation of marine populations is a challenging task relying on information about their current and past abundance. Fisheries‐related data can be scarce and unreliable making them unsuitable for quantitative modelling. One fishery independent method that has attracted attention in this context consists in estimating the effective population size (Ne), a concept founded in population genetics. We reviewed recent empirical studies on Ne and carried out a simulation study to evaluate the feasibility of estimating Ne in large fish populations with the currently available methods. The detailed review of 26 studies found that published empirical Ne values were very similar despite differences in species and total population sizes (N). Genetic simulations for an age‐structured fish population were carried out for a range of population and samples sizes, and Ne was estimated using the Linkage Disequilibrium method. The results showed that already for medium‐sized populations (1 million individuals) and common sample sizes (50 individuals), negative estimates were likely to occur which for real applications is commonly interpreted as indicating very large (infinite) Ne. Moreover, on average, Ne estimates were negatively biased. The simulations further indicated that around 1% of the total number of individuals might have to be sampled to ensure sufficiently precise estimates of Ne. For large marine populations, this implies rather large samples (several thousands to millions of individuals). If however such large samples were to be collected, many more population parameters than only Ne could be estimated.

Population and community indicators for the impact of fishing are often estimated using abundance estimates instead of raw sampling observations. Methods are presented for testing null hypotheses of nonsignificant impacts and, where possible, for calculating the statistical power. The indicators considered concern populations (intrinsic growth rate, total mortality, exploitation rate, and a new indicator, the change in fishing mortality required to reverse population growth) and communities (k- and partial-dominance curves, a biodiversity index, size spectrum, and proportions of various population groups). The performance of these indicators is compared for the Celtic Sea groundfish community based on achieved precision, statistical power, and availability and estimation method of reference points. Among population indicators, mean length of catch was most precisely estimated and the corresponding hypothesis tests had consistently large powers. Total mortality performed reasonably well. In contrast, both the intrinsic population growth rate and the exploitation rate gave unreliable results. All tested community indicators performed similarly well. Indicators for which the direction of change caused by fishing is predictable, such as the proportion of noncommercial species or piscivores in the community, are promising indicators at the community level.

Discarding is an issue of increasing concern and there is a growing number of studies aiming at estimating discard amounts and characteristics. However, the sampling design and methods used in these studies generally rely on implicit assumptions. In this perspective, we examine the available evidence in favour of or refuting these assumptions. We find that (i) the assumptions most commonly used for estimating discards, namely that discards are proportional to catch or to effort, are generally not supported by the available evidence, (ii) both environmental conditions and fishing methods influence the amounts and composition of discards, but because of the huge variability, sampling stratification according to these factors might not result in any improvement of the precision of discard estimates, and (iii) many intricate factors can play a role in determining discards in a particular fishery. We conclude that assumptions should be more carefully checked prior to being taken for granted in discard studies and that more studies designed to improve knowledge of the discarding processes are needed.