The genetic structure of cowpea populations is highly determined by its mating system characterised by a high selfing rate. Moreover, in line with a previous study , our data are also strongly suggestive of genetic exchanges presumably caused by pollinator activity. The genetic variation in Vigna unguiculata ssp unguiculata var. spontanea is low compared to those previously reported in cowpea using allozymes e.g., [15, 30, 31]. However, these studies surveyed several subspecies and were therefore encompassing a much larger part of the cowpea gene pool. Previous results related to var. spontanea (accessions from almost all sub-Saharan Africa)  showed higher diversity than the ones reported here, suggesting that West African var. spontanea represents just a subset of the diversity of the whole var. spontanea. Indeed, the study of Coulibaly et al.  based on AFLP markers (Amplified Fragment Length Polymorphism) markers revealed var. spontanea to be more diverse in eastern than in western Africa. Population genetic study in a wild cowpea population in East Africa (coastal Kenya) shows much higher outcrossing rates than in West Africa (Kouam, unpublished). Reasons explaining the low genetic diversity in the western populations could therefore include a predominantly selfing mating system and/or loss of genetic diversity occurring after genetic bottlenecks during the colonization of dryer savannas linked to the breeding system change .
West African var. spontanea could be classified as primarily selfed plant, according to the criterion of Schemske and Lande . In the present study, outcrossing rates t ranged from 1 to 9.5% across fifteen populations, with a mean equal to 3.4%. The high rates of apparent selfing in wild cowpea populations are consistent with the cowpea flower morphology . In West African var. spontanea, anthers are in contact with the stigmatic surface within the flower bud. Anthers release pollen during the first half of the night  and the cuticle that protects the stigmatic surface is ruptured during the second half of the night, which means that pollen can start to germinate on the stigmatic surface a few hours before the opening of the flower (Pasquet, unpublished observations).
Although consistent with previous results, the outcrossing rates we estimated are markedly higher than previous studies based on pollen flow source and sink trials [21–23]. However, the source and sink trials that were used cannot necessarily detect the shortest pollen moves. Our study focused on naturally occurring populations where individuals (eventually both cultivated and wild plants) can stand few cm apart while source and sink trials typically examine pollen flow between spatially clustered groups of plants usually separated by at least one meter.
A low but significant level of biparental inbreeding confirms the local activity of cowpea pollinators in West Africa. Such trends (low outcrossing rates and low biparental inbreeding) are encountered in numerous wild relatives of inbred legume crops [34–40]. Cowpea pollinators either belong to the genus Xylocopa or the family Megachilidae (Tignegre, unpublished observations). They visit most of cowpea flowers at least once on average. However, these pollinators are expected to do many more flower-to-flower flights within a flower patch than between flower patches [24, 41]. According to Godt and Hamrick , the genetic effect of such pollinator behaviour is to reduce the single-locus outcrossing estimates, as observed here. The rather high level of pollinator activity is counteracted by bud self-fertilization (up to 97%). In turn, this mating system leads to almost complete deviations from Hardy-Weinberg equilibrium with a marked heterozygote deficiency . Moreover, the difference between the observed deficit of heterozygotes (F) and the theoretical equilibrium based on the estimated selfing rates tm
Fe = (1-tm)/(1 + tm)], varied according to the ecological context of the populations. In rather undisturbed habitats the observed inbreeding tended to be higher than the inbreeding equilibrium while this trend was reversed in disturbed habitats (field and roadside). Field and roadside are subjected to frequent disturbances and are characterized notably by ground transfers causing possible rearrangements in the soil seed bank. Moreover, outcrossing rates may vary among years [44, 45]. This could explain why the expected inbreeding equilibrium derived by selfing rate does not necessarily reflect the parental inbreeding.
Because pollen flow is expected to be sharply reduced when distance increases , local gene exchanges should mainly take place within populations, including between wild and domesticated plants when both are mixed or in close proximity. Accordingly, strong genetic differentiation should take place among populations. The existence of spatial genetic structure and its scale of organization is a reflection of gene flow in space and time in relation to the spatial distribution and the colonization history of populations. A spatial genetic structure was found for proximate wild cowpea populations up to 100 km, which reflects a decreased probability to observe related individuals as the distances between populations increase. This suggests genetic exchanges among populations; however, our results do not directly shed light on the patterns of occurrence of gene flow in space and time. Gene flow via pollen in cowpea is likely to occur up to few km with a very low probability of long distance pollen dispersal, and, in any cases, pollen movement is unlikely at distances over 10 km . On the other hand, seed flow through ingestion by grazing mammals could involve much longer distances, though the percentage of seed survival through grazing mammal gut does not exceed a few per cent (Pasquet, unpublished observations). Wild cowpea is expected to express high levels of genetic differentiation and low levels of within-population genetic diversity. In this study, high genetic differentiation was observed at several spatial levels.
Significant allelic correlations between wild and domesticated pairs located in a same zone (distance < 100 km) suggested possibilities of genetic exchanges between these two compartments. Such correlations could arise from multiple local domestications of cowpea. However, considering that cowpea domestication took place more than 3500 years (or generations) ago, and that cv.-gr. Melanophthalmus is the result of two bottlenecks which are likely to have occurred in different places in Africa , an alternative explanation for the allelic correlations must be proposed and is supported by results from Pasquet , Coulibaly , and Feleke , respectively based on allele Amp2
, AFLP alleles, and a chloroplastic DNA. In these studies, alleles characterizing domesticated cowpea accessions, which are rather unfrequent in wild cowpea accessions, could have been useful to locate the center of origin of the crop. Surprisingly, such alleles were found to be widespread across Africa. The authors concluded that the presence of these alleles in wild cowpea accessions was the result of introgression of domesticated alleles into wild gene pool. Our present results represent a more direct evidence of these introgressions into wild cowpea populations.
In an area largely dominated by the cultivation of cv. gr. Melanophthalmus which is characterized by three recessive traits, i.e., white colored seeds, thin and wrinkled seed testa, as well as non-shattering pods , gene flow between wild and domesticated cowpea is expected to be highly asymmetrical . Because the probability of wild dominant alleles entering the domesticated gene pool is almost null, the asymmetry is expected to be higher in cowpea than in other crops where the phenomenon has been observed [39, 47–52]. If a farmer sows a seed from a domesticated flower that has been fertilized by wild pollen, the F1 plant will show shattering pods and smaller seeds with a thick and dark testa. Farmers are not likely to select such small seeds with thick and dark testa for the next sowing, and therefore prevent the introduction of wild alleles into the domesticated genepool in areas where cv.-gr. Melanophthalmus is cultivated exclusively. Of course, pollen from such an F1 plant may fertilize a flower from a domesticated plant, but the probability of recovering the domesticated phenotypes (i.e., finding the combination of recessive alleles in the progeny of such a natural BC plant) is very low in the end. This hypothesis is confirmed for domesticated cowpea by the near-absence of variability in its populations as well as the absence of some of the alleles encountered in wild cowpea (Amp4
and especially Amp2
Because wild and domesticated plants still co-exist, it is likely that positive and negative factors affecting the survival of hybrids balance each other out. Genetic swamping by domesticated genes would lead to the disappearance of wild types, which is obviously not the case. With the exception of one single wild plant, no large-seeded wild cowpea was collected. The exception produces partly white seeds and resulted probably from introgression with domesticated genes. However, this situation appeared to be an exception to a general rule. If gene flow from domesticated to wild cowpea does exist, the lack of strong genetic swamping and modified seed morphology in the wild populations suggests that these introgressions should be rare. Alternatively, gene flow might be rather frequent while hybrids in non-cultivated environment are expected to be less fit. The white seed color makes seeds more visible to seed predators, the thinner seed coat makes seed less dormant, and reduced seed shattering could reduce dispersal distances.