Genetic analysis of natural populations has allowed biologists to ask a wide variety of questions which previously could only be answered by extensive observation of the group in question. A number of genetic markers have proven to be useful. These include mitochondrial DNA, Major Histocompatibility Complex loci, allozyme loci, and Variable Number of Tandem Repeat (VNTR) markers. VNTR markers are characterised by a core sequence which consists of a number of identical repeated sequences. They can be divided into two categories based on the repeat length. These are minisatellite, 15 -70 base pairs (bp), and microsatellites, 2-6 bp.
Recently microsatellites have been increasingly used as the marker of
choice. There are some advantages to utilising microsatellites over
the other markers, which make them desirable. Microsatellite loci are
found in large numbers and are relatively evenly spaced throughout the
genome. For example, Edwards et al.
(1991)
examined the frequency of 5 microsatellite loci (tri
and tetra repeats) on the X chromosome. They found, for either the tri
or tetra microsatellite loci, that any given repeat was found every 300
to 500 Kb. From this they estimated that for all the 44 possible
unique trimeric and tetrameric repeats there are 
400,000 loci or
about 1 every 10 to 20 Kb (Edwards et al.
1991).
Of the class of loci examined by
Edwards et al. (1991)
50% are
polymorphic. Further, most of these loci are selectively neutral which
makes them compatible with the assumptions of most population genetic
theory. Technically microsatellites are more desirable than the larger
VNTR loci because they can be analysed via the Polymerase Chain
Reaction (PCR) (although improvements to the thermal stable polymerases
are enabling the amplification of larger fragments of DNA) and the
alleles can be unambiguously sized on polyacrylamide gels. The lack
of exact size resolution of larger VNTR loci has lead to procedures
such as binning which reduce the statistical power of the analysis
(Budowle et al. 1991). PCR
analysis of small fragments also allows the analysis of degraded
samples in which the mean fragment size of the genomic DNA has been
severely reduced through environmental insult (e.g.
Paabo et al. 1989). Finally,
microsatellites have been found to be variable even in populations
which have low levels of allozyme and mitochondrial variation (Estoup
et al. 1995a,
1995b, 1996;
Paetkau and Strobeck, 1994).
Microsatellites are useful for a number of analyses. They were originally utilised for genetic mapping (e.g. Weissenbach et al. 1992) and have been extensively used for linkage analyses in the association with disease susceptibility genes (e.g. Robinson et al. 1996). In addition they have proven useful in the analysis of paternity and kinship (Queller et al. 1993) and in the probability of sample identity at both the individual (Edwards et al. 1992) and population levels (Paetkau et al. 1995). In the study of entire populations microsatellites are also very useful (Bruford and Wayne 1993). Microsatellite variation has been used to study the amount of hybridisation between closely related species (Gottelli et al. 1994; Roy et al. 1994). Comparison of levels of variation between species and populations have also proven useful in the assessment of overall genetic variation (Gottelli et al. 1994; Paetkau and Strobeck 1994; Taylor et al. 1994). They can be used to estimate effective population size (Allen et al., 1995) and to gain insight into the degree of population substructure including both the amount of migration between subpopulations (Allen et al. 1995; Gottelli et al. 1994;) and genetic relationships among the various subpopulations (e.g. Bowcock et al. 1994; Forbes et al. 1995; Estoup et al. 1996; Lade et al. 1996).
The analysis of population substructure will be the focus of the
following review. Presently there are a wide variety of procedures
which can be utilised to analyse microsatellite data. These include
various estimates of genetic distance between populations, such as
average shared allele distance 
(Chakraborty and Jin
1993;
Bowcock et al.
1994;
for examples of their use with empirical data see
Bowcock et al.
1994;
Estoup et al.
1995a,
1995b,
1996), Chord distance 
(Cavalli-Sforza and Edwards
1967;
for examples of their use with empirical data see Estoup et al.
1995a,
1995b,
1996), Nei's standard genetic distance
(Nei 1984;
for examples of their use with empirical data see
Gottelli et al.
1994;
Taylor et al.
1994;
Scribner et al.
1994;
Paetkau et al.
1995;
Lade et al. 1996), and average sum of squares
of the differences in allele size 
(Goldstein et al.
1995a;
Slatkin
1995;
for examples of their use with empirical data see
Forbes et al.
1995).
In addition various estimators of 
can also be utilised, including 
(Nei 1984; for examples of their use with empirical data see
Roy et al.
1994),
(Weir and Cockerham 1984 for examples of their use with
empirical data see Roy et al.
1994),
and 
(Slatkin
1995
for examples of their use with empirical data see
Allen et al.
1995;
Forbes et al.
1995).
This illustrates the wide variety of treatments microsatellite data has
received to estimate the same parameters, mainly genetic distance and
population substructure. The objective of this review is to explain
the various analyses and to evaluate their differences and strengths.