Function of geophagy as supplementation of micronutrients in Tanzania

Abstract

Geophagy is defined as the practice of eating soil and is observed worldwide. In Tanzania, edible soil sticks called pemba are consumed mainly by pregnant women, but the physiological function of eating pemba has not yet been elucidated. The objectives of the present study were to evaluate the physicochemical properties of edible soil sticks compared with ordinary soils in Tanzania and to evaluate the function of geophagy in terms of micronutrient supply and the adsorption capacity of materials such as toxins. The color of eight pemba samples collected from various markets was reddish or whitish and their shape was generally columnar with an average length, width and weight of 6.1 cm, 1.8 cm and 22 g, respectively. Pemba had a more clayey texture (48% clay on average) than the ordinary soils investigated for comparison, and the clay composition was generally dominated by kaolinite. The concentrations of available nutrients in pemba, extracted with 0.1 mol L⁻¹ NaCl (pH 2), were 391 mg Ca kg⁻¹, 234 mg Mg kg⁻¹, 17 mg Mn kg⁻¹, 6.5 mg Fe kg⁻¹, 4.9 mg Cu kg⁻¹, 2.8 mg Co kg⁻¹, 1.9 mg Zn kg⁻¹ and 1.1 mg Ni kg⁻¹, and extraction with reductant drastically increased the amounts of Fe and Mn released. The amount of available nutrients supplied by pemba consumption at a rate of 50 g day⁻¹ amounted to 99% of Mn and 13% of Fe in the case of reddish pemba and 75% of Cu in the case of whitish pemba of the daily requirement by pregnant women, suggesting the potential of pemba to supply these micronutrients. A moderate cation exchange capacity (CEC) level (11.1 cmolc kg⁻¹) and lower ratio of CEC to clay content for the pemba compared with the soils indicated that the adsorption capacity was not the main criteria for choosing soil materials and instead fine-textured soils with kaolinitic clay mineralogy were deliberately chosen for pemba. In conclusion, the main function of eating pemba in Tanzania, and probably the original function of geophagy, is the supply of micronutrients, such as Mn, Cu and Fe, rather than the adsorption of toxic materials.

Key words:

• geophagy
• kaolinite
• micronutrients
• pemba
• soil

INTRODUCTION

Geophagy is defined as the practice of eating soil and has been observed in many species, including mammals, birds, reptiles and butterflies (Diamond 1999; Kreulen 1985), and humans are no exception. Geophagy by humans has been observed worldwide and is common in children and pregnant women (Halsted 1968; Oliver 1997). There are several hypotheses on the reason why some people eat soils. The first one is to supplement essential minerals, such as iron (Fe), zinc (Zn), copper (Cu), manganese (Mn), calcium (Ca) and magnesium (Mg) (Johns and Duquette 1991a; Moynahan 1979). There have also been some reports, however, that geophagy might increase anemia (Okcuoglu et al. 1966) because it reduces the uptake of Fe (Minnich et al. 1968). The second hypothesis is to relieve gastric complaints and nausea (Vermeer and Ferrell 1985) and to detoxify toxic materials because clay soil can adsorb dietary toxins, such as alkaloids, tannin and other plant constituents (Diamond 1999; Dominy et al. 2004; Johns 1986; Johns and Duquette 1991a,b). Alternatively, geophagy might be carried out to fill the stomach as a famine food or it might be regarded as a custom or cultural activity. Gelfand (1945) speculated that pregnant women eat soil because of the observed soil's fertility, hoping for the future fertility of their children. Thus, there are many explanations for geophagy. However, the function of eating soils has not been fully elucidated because most research has focused on only one aspect of geophagy, that is, supplementation of micronutrients or detoxification of toxic materials, and has not compared other possible functions using the same samples.

Geophagy is widespread in the tropics, especially in Africa (Abrahams and Parsons 1996). In Tanzania, the custom of geophagy is widely observed and there are several types of edible soil sticks, called pemba, that are consumed mainly by pregnant women (according to interviews with the local people). Judging from the fact that geophagy has been observed in Africa for a long time, it would be reasonable to investigate the functions of geophagy in Africa, on the assumption that its original style and function are preserved. Therefore, the objectives of the present study were: (1) to evaluate the physicochemical properties of edible soil sticks compared with ordinary soils in Tanzania, (2) to evaluate the function of geophagy in Tanzania in terms of the supply of micronutrients and the adsoprtion of toxic materials, (3) to classify the functions of geophagy in relation to clay mineralogy with reference to the development of the soil–human relationship.

MATERIALS AND METHODS

Edible soil sticks and soils

Eight types of edible soil sticks (pembas) in a variety of colors and shapes were purchased at local markets in Dar es Salaam (6°51′S, 39°18′E), Morogoro (6°44′S, 37°45′E), Mpanda (6°30′S, 30°45′E) and Kigoma (4°52′S, 29°38′E) in Tanzania. Five to ten samples were collected for each type. These pemba samples were available at ordinary markets and were consumed mainly by pregnant women (according to interviews with the local people). The price was generally one Tanzanian shilling per stick, equivalent to approximately US$0.01 or ¥1 in 2005. We were unable to determine how the pemba were made, but it has been reported that in Nigeria women scrape away coarse and gritty materials, sun-dry the irregular blocks, and after drying, smoke and harden the blocks for a few days (Vermeer and Ferrell 1985). The meaning of the word pemba is not clear, but it might be related to the name of an island approximately 100 km north of Dar es Salaam because similar geophagical clays are called Calabar in Ghana and Togo, a name suggesting connection with the old slave port of Calabar on the Guinea coast (Vermeer and Ferrell 1985). A wide range of colors and shapes of the pemba samples collected from several regions implied that these samples were representative of pemba in Tanzania.

Fourteen soil samples were collected mainly from the plateau area in Tanzania to cover a range of soil types for the comparison of ordinary soil with the edible soil sticks. Eight samples were collected from surface (A) to argillic (Bt) horizons in two profiles, that is, Udult and Ustalf (Soil Survey Staff 2003), and six samples were collected from Bt horizons mainly from Udults. Samples were mainly selected from Bt horizons because pemba was thought to be made from fine-textured soil materials. The samples were air-dried, disaggregated using a porcelain pestle and mortar and sieved to <2 mm before chemical analysis.

Corn samples

Five sets of corn samples were collected at Dar es Salaam and Morogoro as representative samples of staple food in Tanzania. The samples were thoroughly ground to a powder before chemical analysis.

Analysis of the pemba, soil and corn samples

To evaluate the appearance of the pemba samples, their size was measured (length, width and weight) and their color was determined by hue, value and chroma according to the Munsell color system.

The texture and mineralogical characteristics of the pemba and soil samples were measured as follows. The sand, silt and clay contents were determined by the sieving and pipette method (Gee and Bauder 1986). Clay mineralogy was determined by the X-ray diffraction method with CuK_α radiation (MiniFlex, Rigaku, Tokyo, Japan). X-ray diffraction patterns of oriented specimens were obtained for Mg-clay without and with glyceration and for K-clay both before and after heating at 150, 350 and 550°C for 2 h to determine the relative amount of layer silicate minerals. X-ray diffraction patterns of non-oriented specimens were also obtained to determine the relative amount of gibbsite, goethite and hematite. For chemical characteristics, the pH and electrical conductivity (EC) were determined electrochemically. Total C content was determined by the dry combustion method (Sumigraph NC-95A, Sumika Chemical Analysis Service, Osaka, Japan). Exchangeable Ca, Mg, Na and K were determined by atomic absorption spectrometry (AA-6200; Shimadzu, Kyotot, Japan) after extraction with a neutral 1 mol L⁻¹ ammonium acetate solution. The cation exchange capacity (CEC) was determined, as an index of adsorption capacity, by the colorimetrical measurement of NH⁺ _- using an improved indophenol method (Reine et al. 1998) after extraction of the retained NH⁺ _- at pH 7 with 1 mol L⁻¹ NaCl solution.

Figure 1  Appearance of the edible soil sticks (pemba) in Tanzania. From left to right: Pr1, Pr2, Pr3, Pr4, Pr5, Pw1, Pw2 and Pw3. The coin on the right is 2 cm in diameter.

To characterize the nutritional status of the pembas and soil samples, the total and available fractions of eight elements, Fe, Mg, Ca, Mn, Zn, Cu, nickel (Ni) and cobalt (Co), were determined. For the total analysis, finely ground samples were wet digested with concentrated nitric acid, perchloric acid and hydrofruoric acid. For the analysis of the available fraction, 1 g of sample was extracted with 30 mL of 0.1 mol L⁻¹ NaCl solution adjusted to pH 2.0 with HCl for 2 h at room temperature because a pH value of 2 is representative of the normal stomach pH and the use of 0.1 mol L⁻¹ NaCl provides constant ionic strength for all experiments (Johns and Duquette 1991a). In addition, the same extraction method was carried out with a reductant to simulate the possible digestion of pemba under reduced conditions; 1 g of sample was extracted with 30 mL of 0.1 mol L⁻¹ NaCl and 10 g L⁻¹ hydroxlamine chloride solution adjusted to pH 2.0 with HCl for 2 h at room temperature. The concentrations of Fe, Mg, Ca, Mn, Zn, Cu, Ni and Co in the solution were subsequently determined by inductively coupled plasma atomic emission spectrometry ([ICP-AES] SPS-1500VR; Seiko Instruments, Chiba, Japan).

Total analysis of the concentrations of the eight elements in the corn samples was also carried out by wet digesting plant samples and measuring their concentrations by ICP-AES.

RESULTS

Color and shape of the edible soil sticks

Figure 1 shows the appearance of the edible soil sticks and Table 1 provides data on their shape and color. Judging from the color of the sticks, there were two types of pembas: five were reddish (bright reddish brown to orange) and three were whitish (light gray). The reddish pembas were designated, hereafter, as Pr1 to Pr5 and the whitish pembas as Pw1 to Pw3. The color of the pembas is believed to reflect the mineralogy of the soil constituents from which the sticks were made. The shape of the soil sticks was generally columnar, but differed considerably. The length and width of the sticks ranged from 4.1 cm to 8.0 cm and 1.1 cm to 2.3 cm, with mean values of 6.1 cm and 1.8 cm, respectively. Their volume of the pembas, calculated from the length and width, ranged between 6.6 cm³ and 30.3 cm³, with a mean value of 16.3 cm³. The weight of the pemba sticks ranged from 7.4 g to 39.8 g, with a mean value of 22.0 g and their density was approximately 1.3 g cm⁻³, with some variation.

Table 1 Color and shape of the pemba

Sample		Color	Length (cm)	Width (cm)	Volume (cm^3)	Weight (g)	Density (g cm^−3)	Location^†
Pr1	2.5YR5/6	Bright reddish brown	4.8	1.8	11.4	10.4	0.9	Mpanda
Pr2	5YR5/6	Bright reddish brown	8.0	2.0	25.1	39.0	1.6	Dar es Salaam
Pr3	5YR6/8	Orange	7.4	2.3	30.3	39.8	1.3	Kigoma
Pr4	5YR5/6	Bright reddish brown	6.8	1.1	6.8	13.1	1.9	Kigoma
Pr5	5YR5/8	Bright reddish brown	7.2	2.0	23.0	33.0	1.4	Morogoro
Pw1	5Y8/1	Light gray	4.9	1.4	6.9	7.9	1.1	Mpanda
Pw2	7.5Y8/1	Light gray	4.1	1.4	6.6	7.4	1.1	Dar es Salaam
Pw3	7.5YR8/2	Light gray	5.9	2.1	20.2	25.7	1.3	Kigoma
Average (Pr)			6.8	1.8	19.3	27.1	1.4
Average (Pw)			5.0	1.6	27.1	13.7	1.2
Average (Total)			6.1	1.8	16.3	22.0	1.3
^†Location of the local market where the samples were purchased.

Table 2 Texture and mineralogy of the pemba and soil samples

Sample	Sand (%)	Course sand (%)	Fine sand (%)	Silt (%)	Clay (%)	Textural class	Clay mineralogy
Pemba
Pr1	28	12	16	26	46	Heavy clay	Kt>>It>Qz, He
Pr2	29	12	17	17	54	Heavy clay	Kt>>He
Pr3	31	16	15	32	38	Light clay	Kt>>It, Qz, He, Ge
Pr4	31	14	18	30	39	Light clay	Kt>>It>Qz, He
Pr5	34	15	19	27	39	Light clay	Kt>It>>Qz, He, Ge
Pw1	3	2	1	35	62	Heavy clay	Kt
Pw2	8	6	2	32	60	Heavy clay	Kt
Pw3	11	8	2	43	46	Heavy clay	Kt>>It, Qz
Average (Pr)	31	14	17	26	43	Light clay
Average (Pw)	7	5	2	37	56	Heavy clay
Average (Total)	22	11	11	30	48	Heavy clay
Soil
Average	56	41	15	10	33	Sandy clay	Kt>>It, Vt, He, Ge
Maximum	91	72	30	24	69	Heavy clay
Minimum	16	11	5	2	6	Loamy Sand
CV (%)	46	52	55	66	66
CV, coefficients of variation; Ge, goethite; He, hematite; It, illite; Kt, kaolinite. Qz, quartz; Vt, vermiculite.

Comparison of the pemba and soil samples in terms of texture, mineralogy and chemical characteristics

Table 2 shows the texture and mineralogy of the pembas and soil samples. On average, pemba consisted of 22% sand, 30% silt and 48% clay, with the textural class of heavy clay. The Pr samples had a slightly coarser texture than the Pw samples. Judging from the fact that the soil samples consisted of 56% sand and 33% clay on average, it is clear that fine-textured soils were deliberately chosen for the use of pemba, particularly for the whitish pemba. The clay mineralogy of the pemba indicates that kaolinite was the dominant clay, with illite, goethite and hematite as minor components, reflecting the highly weathered conditions in these areas. Between Pr and Pw, Pr had much more Fe oxides, particularly hematite, than Pw, as their color implied. The clay mineralogy of the soil samples was generally similar to that of the pemba. These results suggested that color is an indicator of the texture and mineralogy of the soils used for the pemba. The chemical characteristics of the pemba and soil samples are shown in Table 3. The pH of the pemba was approximately 5, similar to that of the soil, reflecting the highly weathered conditions. The total C content was approximately 2.5 g kg⁻¹, and slightly lower than the average recorded for the soils, possibly because of the fact that the soils used for the pemba were collected from B horizons where clay tended to accumulate and organic matter tended to be lower. The CEC of the pemba was on average 11.1 cmolc kg⁻¹. The CEC of the Pr samples (12.5 cmolc kg⁻¹ on average) was slightly higher than that of the Pw samples (8.7 cmolc kg⁻¹). These values were comparable to the values recorded for the soils (13.3 cmolc kg⁻¹ on average). The EC and exchangeable cations showed similar tendencies for the pemba and soil samples.

Table 3 Chemical characteristics of the pemba and soil samples

Sample	pH	EC (µS cm^−1)	Total C (g kg^−1)	Exchangeable bases				CEC
				Ca	Mg	Na	K
				(cmolc kg^−1)
Pemba
Pr1	5.0	89	3.8	1.04	1.33	0.23	0.28	12.6
Pr2	5.4	54	0.6	3.70	1.58	0.16	0.11	9.5
Pr3	4.9	41	2.0	0.92	2.52	0.09	0.19	13.1
Pr4	4.8	124	5.0	1.23	3.76	0.84	0.27	13.8
Pr5	4.7	200	0.2	1.23	2.33	0.83	0.29	13.5
Pw1	5.4	78	3.3	6.01	4.41	0.33	0.20	10.5
Pw2	4.8	58	2.9	2.99	1.71	0.10	0.13	8.0
Pw3	4.1	230	2.4	2.28	1.46	0.13	0.13	7.6
Average (Pr)	5.0	102	2.3	1.63	2.30	0.43	0.23	12.5
Average (Pw)	4.8	122	2.8	3.76	2.52	0.19	0.15	8.7
Average (Total)	4.9	109	2.5	2.43	2.39	0.34	0.20	11.1
Soil
Average	5.2	80	8.1	2.46	1.80	0.03	0.79	13.3
Maximum	5.4	210	35.9	11.57	11.68	0.10	2.71	33.5
Minimum	5.0	26	2.1	0.07	0.12	0.01	0.00	2.1
CV (%)	2	72	107	140	163	81	136	72
CEC, cation exchange capacity; CV, coefficients of variation; EC, electrical conductivity.

Table 4 Total concentration of eight nutrients in the pemba

Sample	Fe	Mg	Mn	Ca	Zn	Cu	Co	Ni
	(g kg^−1)		(mg kg^−1)
Pemba
Pr1	127	2.9	683	332	64	95.3	39.8	20
Pr2	98	0.5	135	774	56	26.2	16.4	40
Pr3	147	4.0	994	345	78	118.2	62.9	23
Pr4	151	4.9	1107	429	89	84.0	64.5	23
Pr5	153	4.1	1097	400	80	70.2	65.0	21
Pw1	15	0.7	173	684	54	101.8	46.1	80
Pw2	17	0.4	90	662	20	82.3	30.9	33
Pw3	24	0.5	57	513	242	41.5	16.4	98
Average (Pr)	135	3.3	803	456	73	79	50	26
Average (Pw)	19	0.6	106	620	105	75	31	70
Average (Total)	92	2.3	542	517	85	77	43	42

Nutritional status of the pemba

Table 4 shows the total concentrations of eight nutrients in the pemba. On average, the order of the total content was as follows: Fe (92 g kg⁻¹) > Mg (2.3 g kg⁻¹) > Mn (542 mg kg⁻¹) > Ca (517 mg kg⁻¹) > Zn (85 mg kg⁻¹) > Cu (77 mg kg⁻¹) > Co (43 mg kg⁻¹) > Ni (42 mg kg⁻¹). It should be noted that the contents of Fe, Mg, Mn and Co were relatively higher in Pr, whereas the contents of Ca, Zn and Ni were relatively higher in Pw. These data are believed to indicate the potential of pemba to supply these nutrients, although most of these elements would not be readily available.

Table 5 Concentration of the available fraction of eight nutrients in the pemba

Sample	Fe	Mg	Mn	Ca	Zn	Cu	Co	Ni
	(mg kg^−1)
Without reductant
Pr2	5.1	148	11	846	1.1	0.7	0.9	0.1
Pr3	7.2	275	27	106	2.0	4.3	5.3	1.0
Pr4	7.1	413	33	169	3.0	3.2	4.9	0.7
Pr5	7.2	260	20	283	2.7	2.7	2.8	0.5
Pw2	6.9	165	9	603	1.1	16.8	1.5	1.8
Pw3	5.3	142	3	338	1.5	1.6	1.2	2.6
Average (Pr)	6.7	274	23	351	2.2	2.7	3.5	0.6
Average (Pw)	6.1	153	6	471	1.3	9.2	1.4	2.2
Average (Total)	6.5	234	17	391	1.9	4.9	2.8	1.1
A/T (Pr)^†	0.00	8.4	2.8	76.9	3.0	3.5	7.0	2.2
A/T (Pw)^†	0.03	27.6	5.6	76.0	1.2	12.3	4.4	3.1
A/T (Total)^†	0.01	10.4	3.2	75.6	2.2	6.3	6.5	2.6
With reductant
Pr2	56.2	238	29	1294	1.4	0.9	2.8	0.0
Pr3	121.6	393	185	134	2.2	5.4	13.7	0.9
Pr4	72.5	551	103	209	3.5	3.7	10.9	0.4
Pr5	68.2	305	146	272	3.5	3.7	4.8	0.6
Pw2	34.2	195	30	581	0.9	15.4	3.7	2.0
Pw3	17.1	156	3	335	1.0	2.2	1.4	2.7
Average (Pr)	79.6	372	116	477	2.7	3.4	8.0	0.4
Average (Pw)	25.7	176	17	458	0.9	8.8	2.6	2.3
Average (Total)	61.6	306	83	471	2.1	5.2	6.2	1.1
A/T (Pr)^†	0.06	11.3	14.4	104.7	3.6	4.4	16.2	1.7
A/T (Pw)^†	0.14	31.6	15.6	73.9	0.9	11.7	8.2	3.3
A/T (Total)^†	0.07	13.6	15.2	91.0	2.4	6.7	14.6	2.5
^†Relative ratio of the available fraction to the total. Note: samples Pr1 and Pw1 could not be investigated because of their limited amounts.

The concentrations of the available fractions of the eight nutrients are shown in Table 5. The order of extraction was different from the order of the total concentration, that is, Ca (391 mg kg⁻¹) > Mg (234 mg kg⁻¹) > Mn (17 mg kg⁻¹) > Fe (6.5 mg kg⁻¹) > Cu (4.9 mg kg⁻¹) > Co (2.8 mg kg⁻¹) > Zn (1.9 mg kg⁻¹) > Ni (1.1 mg kg⁻¹). Accordingly, the relative ratio of the available fraction to the total fraction (A/T) varied considerably. Calcium had an extremely high ratio of 75.6% and Mg, Co, Cu, Mn, Ni and Zn had moderate ratios of extractability, 10.4, 6.5, 6.3, 3.2, 2.6 and 2.2%, respectively. In contrast, Fe had an extremely low ratio of 0.01%. Between Pr and Pw, there was a tendency for the contents of Fe, Mg, Mn, Zn and Co to be relatively higher in Pr, whereas the contents of Ca, Cu and Ni were relatively higher in Pw, reflecting their clay mineralogy. These data were comparable to the reported data obtained using the same extraction procedure for edible clays collected from Cameroon, Gabon, Kenya, Nigeria, Togo, Zambia and Zaire (Johns and Duquette 1991a).

Table 5 also shows the effect of a reductant on the extraction of the micronutrients. The reductant drastically increased the amount of Fe and Mn extracted. A similar tendency was also observed for Co. This must be because Fe, Mn and Co occluded in the oxides were released by the reduction.

DISCUSSION

Relationship between the amount of nutrients supplied by pemba and the daily requirement of pregnant women

Table 6 shows the relationship between the daily intake of nutrients through the consumption of pemba and the daily requirement of nutrients of pregnant women. The daily intake of nutrients was calculated based on the assumption that pregnant women consume pemba at a rate of 50 g day⁻¹ or approximately two sticks per day, using the data on available nutrients extracted without or with the reductant. This is because the women are said to consume three sticks per day (according to interviews with local people) and because a rate of 30–50 g day⁻¹ has been reported in Ghana (Vermeer 1971). For the daily requirement, estimated average requirements (EAR) were mainly used, and adequate intake (AI) was used if EAR was not available for the nutrient (World Health Organization 1996). It should be noted that the daily requirements of Mg, Cu and, in particular, Fe of pregnant women are higher than the levels required by non-pregnant women. In addition, the daily intake through corn as a staple food was calculated as a reference, based on the assumption that women eat corn at a rate of 300 g day⁻¹. Based on these data, the relative percentages of the amount of elements absorbed to the daily requirement of pregnant women are also shown on the lower half of Table 6.

Table 6 Relationship between the daily intake of available nutrients through pemba or corn flour and the daily requirement

	Fe	Mg	Mn	Ca	Zn	Cu	Co	Ni
	mg day^−1 (%)^‡‡
Intake-A (Pr)^†	0.33 (2)	13.7 (5)	1.1 (33)	17.5 (3)	0.11 (2)	0.14 (23)	0.17	0.03
Intake-B (Pr)^‡	3.98 (24)	18.6 (7)	5.8 (165)	23.9 (3)	0.13 (2)	0.17 (29)	0.40	0.02
Intake-A (Pw)^†	0.30 (2)	7.7 (3)	0.3 (8)	23.5 (3)	0.06 (1)	0.46 (77)	0.07	0.11
Intake-B (Pw)^‡	1.28 (8)	8.8 (3)	0.8 (24)	22.9 (3)	0.05 (1)	0.44 (73)	0.13	0.12
Intake-C (corn)^§	3.17 (19)	124 (48)	0.8 (22)	5.9 (1)	2.30 (38)	0.78 (130)	0.00	0.09
Daily requirement X^¶	16.5^§§	260^§§	3.5^¶¶	700^¶¶	6^§§	0.6^§§
Daily requirement Y^††	9.0^§§	230^§§	3.5^¶¶	700^¶¶	6^§§	0.5^§§
^†Intake (50 g day^−1) based on extraction without reductant.
^‡Intake (50 g day^−1) based on extraction with reductant.
^§Intake (300 g day^−1).
^¶Daily requirement by pregnant women.
^††Daily requirement by non-pregnant women.
^‡‡Relative percentage to the daily requirement by pregnant women (daily requirement A).
^§§Estimated average requirements.
^¶¶Adequate intake.

Among the nutrients investigated, the daily intake of Fe, Mn and Cu through pemba was relatively high, that is 2–24% for Fe, 8–165% for Mn and 23–77% for Cu. It is also clear that the reduced condition, which is often seen in digestive organs, could greatly increase the amount of Fe and Mn absorbed, but not the amount of Cu absorbed. This is because extraction with the reductant greatly increased the amount of Fe and Mn released (Table 5). In relation to the type of pemba, Pr tended to have a greater potential for the supply of Fe and Mn, particularly under the reduced condition, and Pw had a greater potential for the supply of Cu, irrespective of the reduced condition, possibly reflecting their mineralogical characteristics. Accordingly, the consumption of reddish and whitish pemba could contribute 99% of Mn and 13% of Fe, and 75% of Cu, respectively, of the daily requirement of pregnant women. In contrast, the relative contribution of the Na requirement of pregnant women (EAR: 600 mg day⁻¹) was less than 1% for both Pr and Pw (based on the data of exchangeable Na), suggesting that Na is not a nutrient that is effectively absorbed by pemba consumption.

The daily intake of Fe, Mn and Cu through corn consumption was 3.2, 0.8 and 0.8 mg day⁻¹, that is, 19, 22 and 130% of the daily requirement of pregnant women, respectively. Therefore, Mn is the nutrient most influenced by the consumption of pemba in terms of the daily requirements of pregnant women because the Mn supply by pemba corresponded to a considerable portion (57%) of the daily requirement and was much higher than the supply in corn (22%). Copper and Fe are also likely to be affected by the consumption of pemba. In contrast, the daily intake of Mg, Ca and Zn by pemba consumption was much less than 10% of the daily requirement of pregnant women, suggesting that it is unlikely that pemba can be used to supply these elements. With regard to Co and Ni, there were no reliable data on the daily intake by pregnant women, so it was not possible to evaluate the function of pemba in supplying these elements. Pemba might contribute to the daily intake of Co considerably, however, judging from the fact that there is almost no Co in corn. In conclusion, the potential of pemba to supply micronutrients was clearly demonstrated, particularly for Mn, Fe, Cu and possibly Co. Further investigations of the concentrations of micronutrients in other food stuffs, such as meat, fish and other vegetables, should be conducted to provide an entire picture of the nutritional status of people in this region.

Adsorption capacity of pemba

Many reports have suggested that the adsorption of toxic materials is the main function of geophagy (Diamond 1999; Dominy et al. 2004; Johns 1986; Johns and Duquette 1991a,b). Thus, the adsorption capacity of pemba was investigated. Figure 2 shows the relationship between the clay content and the CEC, an index of adsorption capacity, for the pemba and soil samples in Tanzania. The CEC of the pemba was approximately 10 cmolc kg⁻¹, which was similar to the average soil CEC, 13.3 cmolc kg⁻¹ (Table 3). It is clear that the soils used to make pemba were not chosen because they have a higher CEC or higher absorption capacity, since the soil CEC varied from 2.1 to 33.5 cmolc kg⁻¹. This is in contrast to findings showing that soils with a high CEC are consumed in Turkey, which might be related to anemia resulting from limited Fe absorption (Minnich et al. 1968), and that parrots in Peruvian Amazon rainforest prefer soils with a much higher CEC than an adjacent band of rejected soils, probably to detoxify the birds’ plant diet (Diamond 1999). In contrast, the ratio of CEC to clay content indicates that the ratio of pemba was lower than that of soils: on average 0.16 for Pw, 0.29 for Pr and 0.41 for soils. This result suggests that the soils used for pemba were chosen to have a lower ratio, that is, a relatively high clay content and relatively low CEC. People deliberately choose heavy-textured and kaolinitic soils to make pemba. This might be because kaolinite does not usually interfere with the absorption of Zn or Fe because of its low CEC (<10 cmolc kg⁻¹), whereas minerals with a higher CEC interfere with the absorption of Fe (Minnich et al. 1968), or because kaolinitic clay serves as a traditional antidiarrheal pharmaceutical (Vermeer and Ferrell 1985). Thus, the adsorption capacity of pemba was not prominent and the adsorption of toxic materials is not the main reason why pemba is consumed by pregnant women in Tanzania. The selection of relatively low CEC kaolinitic soils supports the hypothesis that the supply of micronutrients is the main function of eating pemba.

Figure 2  Relationship between the clay content and the cation exchange capacity (CEC) for the pemba (Pr; •), pemba (Pw; ▴) and the soil (□) samples.

Conclusions

According to the results obtained in the present study, it is possible to conclude that the clay mineralogy of the soil consumed is the key factor determining the main function of geophagy. The supply of micronutrients, such as Mn, Fe and Cu, appears to be the main function of geophagy if the clay mineral is dominated by kaolinite or 1:1 type clay with a relatively low adsorption capacity. The adsorption of toxic materials might be the main function of geophagy when the clay mineralogy is dominated by 2:1 type clay with a relatively high adsorption capacity.

Judging from the fact that Homo sapiens originated from Africa approximately 200 thousand years ago (Cann et al. 1987), and spread to other continents approximately 80 thousand years ago (Oppenheimer 2003), and that geophagy is widely observed in Africa at present (Abrahams and Parsons 1996, 1997), it is reasonable to assume that geophagy originated in Africa using African soils. Because soils in Africa are mainly dominated by highly weathered soils with a kaolinitic clay mineralogy (Food and Agriculture Organization–UNESCO 1977), it can be speculated, therefore, that the original function of geophagy was to supply micronutrients under constant malnutrition. The spread of the custom of geophagy worldwide to other continents with a variety of soils with different clay mineralogy might have resulted in additional functions of geophagy in a range of situations. In this context, further investigations into the background of this habit are needed to understand local situations in more detail. Finally, it can be concluded that geophagy reflects the most primitive relationship between human beings and soil, which originated far earlier than another soil–human relationship, that is, agriculture.

ACKNOWLEDGMENTS

The authors wish to thank Dr Shinya Funakawa, Dr Tetsuhiro Watanabe and Mr Tetsushi Yoshida, Kyoto University, for their valuable suggestions and assistance throughout this experiment.