Previous works and early exploration have shown that B. sapida is distributed in different phytogeographic zones of Benin. Each phytogeographic zone hosts various ethnic groups and members of the same ethnic group are sometimes dispersed across different phytogeographic zones historically. However, even if people belonging to the same ethnic group are settled in different locations, they share together traditions, historical experiences, perceptions, values, attitudes, beliefs and language. Therefore, one may expect some variability on uses of natural resources and subsequent know-how not only among ethnic groups, but also among gender group.

According to the above-mentioned considerations and in order to get the maximum of information, eleven communes distributed in the three main phytogeographic zones (Figure 1) where B. sapida is known and used by local populations were included in the survey. In each commune, between 20 or 30 persons were randomly chosen. In total, 240 persons (50% women) belonging to the following Beninese ethnic groups (Adam and Boko, 1993): Batombu, Yoruba, Otamari, Natemba and Fon (Table 1). In addition, 5 women active in the processing of ackee fruits and 6 traditional healers were included.

Semi-structured interviews concerning the species were carried out once. Information's collected refer to the denomination of the species and its meaning, the motivation of the respondents to conserve ackee trees in their land use system, local uses, perception of variation, preferences in fruits traits, management practices to improve the production and regenerate ackee, the gender-specific tasks and responsibilities in the production and processing of ackee products.

The different traditional products obtained from ackee trees and inherent processing techniques were recorded. Likewise, the processing steps of each product, their variability and subsequent constraints were also recorded.

Frequency distribution was used to compare answers within each ethnic group. The fidelity level (FL) 12 was calculated using the following formula: FL = Ip/Iu × 100%, where Ip is the number of informants answering positively on each question and Iu the total number of positive answers for each category of the questionnaire. This index was used to measure the consensus degree between informants and the relative importance of each category of knowledge within each ethnic group.

The interviewees were grouped according to ethnic group and gender group (men and women) so that in each ethnic group, two subgroups were defined: men (M) and women (F). With five ethnic groups, 10 subgroups were constituted. Because the size of subgroups differed and an interviewee could choose more than one ackee trait, the relative frequency of each trait was determined for each of the 10 subgroups. This parameter is defined as the proportion of interviewees belonging to the subgroup who identified the particular ackee trait. A data matrix comprising the relative frequencies of ackee traits was then submitted to Principal Component Analyses (PCA) using STATISTICA 8.0 13 . This statistical method was used to identify traits that best explained the pattern of variation according to the different subgroups. For graphical purposes, the subgroups are labelled by preceding the ethnic group prefix (first three letters) with the label of one of the 10 subgroups defined above. For example, a man from Batombu ethnic group is labelled BatM, whereas a woman from the same ethnic group is labelled BatF.

Loglinear analysis was also performed using the PROC CATMOD available in SAS 14 with gender group (men or women) and ethnic group as dependent variables for each category of answer to detect possible association between people knowledge and their ethnic or gender membership.

B. sapida is designated in each language by different local names shown in Table 1. The name Foulama used by the ethnic group Natemba means "groundnut of trees" by comparison of arils to nuts of peanut (Arachis hypogaea L.). All others local names do not have any particular meaning.

In general, ackee trees are integrated in different land use systems across the country for a variety of reasons including the direct uses as food, soap, medicine, shade, myth and for its marketing value. Apart from the use as food, it was always the combination of two or three other reasons that determined the conservation of ackee in farmers' field. Table 2 shows the percentage of persons quoting each type of motivation in each ethnic group. The main motivation is always the use as food (between 53.3% among the Yoruba and 100% among the Otamari). The Otamari ethnic group showed also the highest motivation frequency for medicinal (73.3%) and marketing (36.7%) reasons. The Natemba is the second group using ackee for its marketing value. Natemba (40%) and Yoruba (30%) are the two ethnic groups valorising ackee soap while the tree provided shade for nearly 19% of the respondent in the Fon group.

In addition, women conserve ackee for soap making and its commercial value, while men keep them for shade. The trade of ackee products seems to be restricted to the ethnic groups Otamari and Natemba. The motivation to conserve ackee trees varied significantly among ethnic groups (χ² = 14.49, df = 4, p < 0.01) but not among gender group (Table 3). From one ethnic group to the other the motivation depended on the gender (χ² = 13.11, df = 4, p < 0.05).

The fidelity level (FL) of motivation highlighted the uses as food, medicine, soap and the commercial value as the most important (Table 2).

The result of the principal component analysis (PCA) performed on the indigenous knowledge, the perception of variation and the traditional management of ackee showed that the first three axes explained 72.8% of the variation observed. Therefore, only the first three axes were used to describe the relationship between people's knowledge of the species and their ethnic group and gender. Table 8 shows the sign of correlation values between the different criteria and the three PCA axes. Figure 3A and 3B shows the projection of the different ethnic/gender groups onto the first and second, first and third axes respectively.

It can be deduced from results shown in Table 5 and Figure 3 that the male Batombu and Otamari (BatM and OtaM) are motivated to conserve ackee on their land for its utility as medicine and myth; use of the dried, boiled or fried aril; favour the use of the soap and as repellent; they perceived variation in fruit size and shape, seed size and aril size and taste; selected fruits based on their size; practiced sowing, assisted tree regeneration, used ackee for mulching and organic fertilization, and practiced pruning and fire protection. The criteria mentioned above are all relevant also for the Fon (men and women) with the following additional ones: market, shade, soap, food and myth in the motivation category; use in fishery; transplantation and tree/crop association (Table 8). The PCA 2 axe clearly separated also the Otamari and Natemba women groups. The Natemba women, the Yoruba women and the Otamari men are highly correlated with the PCA 3 that is explained by the motivation to conserve ackee tree for its utility as soap, the use of the leaves as vegetable, the perception of variation in fruit shape, aril size and aril taste, and the following preferences in ackee fruits traits: fruit shape, aril colour, aril size and aril taste.

B. sapida is well known in Benin. It has been utilized for centuries and is still an important plant genetic resource today. Each ethnic group has different names for the species indicating age-old knowledge and uses. In general, by providing different products, services and having a commercial value, ackee is conserved for its multipurpose properties.

Apart from the use as vegetable in the Natemba ethnic group mentioned above, others utilizations of the aril have been reported also in West Africa 15 16 17 18 and in the Caribbean 19 20 . Farmers managed the species firstly to meet their own needs and also for the commercial value of arils and soap 3 4 .

Future researches on the technological characteristics of the fruits and the design of small storage techniques are needed to help farmers to improve the conservation and long-term storage of arils. The best way to do that would be to establish small processing units managed by farmers in villages or at the communal level. The exact knowledge of composition is the basis for any successful technological process 21 . Preliminary analyses of the physical composition of arils from Toukountouna (North-West Benin) have shown that it includes 46% of oil, 47% of fibres and 3% of proteins 10 . The food value of 100 g of raw arils from Mexico is as follows: Moisture (57.60 g), Protein (8.75 g), Fat (18.78 g), Fibre (3.45 g), Carbohydrates (9.55 g), Ash (1.87 g), Calcium (83 mg), Phosphorus (98 mg), Iron (5.52 mg), Thiamine (0.10 mg), Riboflavin (0.18 mg), Niacin (3.74 mg) and Ascorbic Acid (65 mg) 19 .

Characterization of ackee seed oil and defatted cake of seeds from Southern Benin compared to seeds from Nigeria and Côte d'Ivoire revealed differences in chemical properties and composition of seeds oils (saponification value, iodine value, oleic acidity, peroxide value); fatty acid composition, proximate composition (moisture, fat, crude fibre, total sugars, starch, proteins) and mineral composition (K, N, P, Ca, Mg, Na, Cu, Zn, Mn, Fe, Fe, Ash) and 17 amino acids 22 23 24 . Those results highlighted the potential of ackee seeds for industrial use as lubricants, surfactants and as oil for consumption that should be further explored. Feeding trial experiments conducted in savannas areas of Nigeria have shown that ackee leaves are good fodder resource for West African Dwarf goats especially in the dry season 25 . This is probably good news for animal breeders in the region because ackee trees flush during dry seasons in many part of West Africa when the availability of grasses to feed ruminants decreases drastically.

Some of the medicinal values attributed to ackee in Benin are known in other countries where the species occurs as well. The bark is used in Ivory Coast together with some spices to relieve pains; leaves and barks are used in association to treat sore stomach, epilepsy and yellow fever in Columbia 26 . In Ghana, the bark is one of the ingredients used in a concoction administered for epilepsy; leave juice is used for washing or as drops for sore eyes, conjunctivitis and trachoma; the pulp of twiggy leaves is applied on the forehead to treat migraine/headache 26 . B. sapida is also a natural source of carboxycyclopropylglycine used in pharmacy. The extraction of this non-proteinogenic amino acid from ackee offers the possibility of avoiding the expensive synthetic procedures 27 . Furthermore, B. sapida has antidiabetic activity 28 . The use of this important traditional medicinal knowledge in a rational way remains a challenge to modern scientific disciplines such as pharmacology. More research is necessary to analyze the properties and therapeutic virtues attributed to the soap as a preliminary step to the mechanization of the production. The confirmation of the virtues of ackee soap can boost its production and contribute to a better valorisation of the enormous quantity of capsules and seeds that are usually thrown away.

Ackee is not well known in Benin for its utility as repellent. However, experiments conducted in Trinidad and Tobago had shown that other fruit parts (epidermis, aril and seed) have repellent properties against stored-product insect pests, namely, Callosobruchus maculatus, Cryptolestes ferrugineus, Tribolium castaneum and Sitophilus zeamais 29 30 .

The various differentiation criteria of ackee trees mentioned above calls for an appropriate characterization to investigate the existence of eventual varieties, and/or to detect the effect of the ongoing domestication process practiced by farmers in their different land-use systems. For instance, the positive correlation reported between fruit size and aril size needs to be tested by morphometric study to characterize the diversity of ackee fruits traits.

All traditional silvicultural management practices to improve the production were reported for other important agroforestry parkland species such as Vitellaria paradoxa and Parkia biglobosa (Jacq.) R.Br. ex Benth. in West Africa 31 . There is a consensus among local people that these management practices favour the abundance of better phenotypes, in this case fruits with preferred traits. This link between management techniques and the perception of variation suggests that there is selection going on with a tendency to increase phenotypes producing desired fruits in managed populations. However, reasons for the "superior" perception of trees under management need further analyses. It is particularly challenging and important to distinguish genetic from environment impacts on phenotypes.

Evidence that farmers have domesticated other African indigenous trees has been reported for Dacryodes edulis H.J. Lam and Irvingia gabonensis (Aubry-Lecomte ex O'Rorke) Baill. 32 33 34 ; Vitellaria paradoxa C.F. Gaertn. 35 and Sclerocarya birrea (A. Rich.) Hochst. subsp. caffra (Sond.) Kokwaro 36 . Processes of plants domestication associated to silvicultural management were also documented for many species in the Tehuacán Valley of Mexico including Stenocereus stellatus (Pfeiff.) Riccob. 37 and Ceiba aesculifolia (Kunth) Britten & Baker 38 .

In general, the ethnobotanical survey revealed clearly that indigenous knowledge about ackee varies according to ethnic group and gender. Particularly three ethnic groups (Batombu, Otamari and Natemba) had a great knowledge about the species. In addition, the multivariate analysis showed also clearly the separation between the knowledge of men and women within the Otamari, Batombu and Yoruba communities.

Selection or breeding programs should focus on ackee trees with preferred traits important for local populations. For instance when looking at preferred fruit traits, this study showed that ackee fruit size is the most important trait for men (Batombu, Otamari and Yoruba). Fruit shape, aril colour, aril size and aril taste were the preferred fruit traits for Batombu women, Otamari men and all Yoruba and Natemba. Those differences needs to be taken into account in any research/development programs related to germplasm sampling and ackee improvement.

Domestication can reduce the genetic diversity of wild populations if cultivars replace autochthonous populations on a large scale. It can also increase the level of variability at desired traits in semi-domesticated populations suggesting that varieties may have multiple origins 1 34 . The effect of the artificial selection reported in this study on the genetic diversity and structure of ackee is not yet known and needs to be evaluated properly. This prerequisite is essential to avoid that the intensification of the domestication process will lead to a progressive elimination of individuals with non-desired quality, and a subsequent loss of genetic diversity.

Conservation of ackee genetic resources can be done effectively through cultivation of the species in agroforestry systems, its maintenance on protected areas where they occurs and maintenance of seeds in gene banks. Since preferred traits vary among ethnic groups and gender, the strategy should be specific and should target not only the morphotypes recognized by local populations but should also integrate the population genetics information.