The state of variability within and between populations can be determined by application of both biometrical and molecular procedures. Whilst the former can describe a population in terms of means and variances and the underlying mode of gene action, if an appropriate biometrical design is applied as part of the characterization of a population (Kearsey, 1993), it is only the procedures of molecular genetics which allow a measure of population structure at the level of the gene and genome to be gained. Various molecular tools are available for the characterization of populations. The most widely used to date is electrophoresis of isozymes, but direct DNA methods, including restriction fragment length polymorphisms (RFLPs), randomly amplified polymorphic DNA and amplified fragment length polymorphism (AFLP), are now being exploited. The application of these procedures allows several different parameters of variability within and between populations to be determined. These include the percentage of polymorphic loci, average and effective numbers of alleles per locus, heterogeneity and heterozygosity indices and the various measures based upon the F statistics of Wright (1965).
More recently the application of DNA-based technologies, particularly 'fingerprinting', has provided a wealth of information on the diversity of wild populations and the relatedness of cultivars. For wild species these techniques have been applied to answering specific questions particularly in relation to breeding systems. A comparison of Plantago spp. using RFLP analysis revealed by the M13 probe showed that the inbreeding P. major possessed little variation within populations but marked differentiation between populations whilst the outbreeding P. lanceolata possessed high variability within populations but only moderate variability between populations. Similarly, in an analysis of cultivar differentiation in three species of Bermuda grass (Cynodon), using DNA amplification fingerprinting (DAF), were able to distinguish some closely related cultivars. On the basis of this technique's discriminatory power they recommend that it be used as a method for seed certification and registration purposes. These are just two examples of the many that have been carried out on a wide range of species exploiting the 'fingerprinting' capacity of the molecular methods.
In the majority of applications of these DNA-based technologies comparisons are based upon statistical analyses of the number of 'bands shared'. Such a procedure can be fraught with problems both on a technical and a genetic level. Some of the technical problems and other sources of error which can be encountered in the preparation, running and reading of gels have been considered by Weising et al. (1995), who emphasize the need for technical care and caution in scoring closely spaced bands. In addition there is also the question as to whether two DNA fragments which migrate to a common position on a gel are homologous. This may only be ascertained by extraction of the bands and comparative cross-hybridization. In RAPD analysis of the Lolium/ Festuca complex of species it was found that four out of six amplification products were homologous when tested by Southern hybridization. Although this is a small sample and ranged across genetically diverse species, it does emphasize the need for caution in assuming homology.
The use of individual isozyme loci and similar genetic markers, as measures of variability, is limited in that they give no indication of the genomic associations that are of importance in the maintenance of co-adapted gene complexes. It is only by looking at combinations of markers that such information can be ascertained. In Avena spp. and Hordeum spp., for example, Allard and his coworkers (Allard, 1990; Allard et al., 1993) have shown by comparisons involving up to 14 discrete loci that populations are made up of individuals containing differing multilocus assemblages of favourable epistatic combinations of alleles. These have arisen by rare outcrossing followed by inbreeding to near homozygosity. In A. hirtula, for example, a majority of Spanish populations were found to be polymorphic for different multilocus genotypes, which suggests that 'interactions at the interplant level may contribute to adaptive significance' (Allard et al., 1993). In addition, in this species, polyploidy is present which, in its own right, can lead to a greater allelic diversity and potential for differing multilocus associations. Here again the breeding system reinforces the stabilization of such associations by restricting the degree of recombination that takes place.
The current procedures using molecular markers for assessing population differentiation consider both expressed and non-expressed parts of the genome. Given the ease with which genetic maps may now be constructed, it seems likely that, in future, measures of population differentiation will take into account genome organization and will target those regions of interest. Already this problem is being addressed as a means of determining identity by descent as part of the statutory procedures in determining 'essential derivation' of cultivars (Dillmann et al., 1995).
Recent developments in QTL analysis of crop plants in defined populations, such as F2s and recombinant inbred lines, are now bringing together the power of genome analysis at the molecular level and biometrical procedures which will eventually allow a more detailed understanding of the genetic architecture of traits of both agronomic interest and of importance to evolutionary fitness.
These various studies of population structure provide an insight into the manner in which variability is distributed within a species and some of the factors controlling that pattern. From a conservation aspect it is now necessary to consider how these mechanisms interact to determine the spatial distribution of variability and their implications for collection and conservation.