ArticlePDF Available

Reproductive Immunosuppression and Diet: An Evolutionary Perspective on Pregnancy Sickness and Meat Consumption


Abstract and Figures

Pregnancy sickness, a suite of "symptoms" that frequently co-occur during pregnancy, may be an adaptation providing behavioral prophylaxis against infection. Maternal immunosupression, necessary for tolerance of the fetus, results in gestational vulnerability to pathogens. Throughout the period of maximal vulnerability, dietary behavior is significantly altered via changes in nausea susceptibility and olfaction and the development of marked aversions and cravings. Of food types, meat is both the most likely to carry pathogens and the principal target of gestational aversions and pregnancy taboos. Because meat was prominent in ancestral human diets but hygienic procedures that effectively eliminate the risk of meat-borne infection are recent, such pathogens likely constituted a source of selective pressure on pregnant females throughout human history. Both the relatively low protein and energy demands of the first trimester and the existense of nonmeat alternatives would have allowed for the evolution of time-limited gestational meat-avoidance mechanisms.Complementing these mechanisms, gestational cravings target substances that may influence immune functioning and affect the availability of iron in the gastro-intestinal tract, thereby limiting the proliferation of iron-dependent pathogens. Clinical and ethnographic findings are examined in light of these proposals, and directions for future research are outlined.
Content may be subject to copyright.
Current Anthropology Volume 43, Number 1, February 2002
2002 by The Wenner-Gren Foundation for Anthropological Research. All rights reserved 0011-3204/2002/4301-0002$3.50
and Diet
An Evolutionary Perspective on
Pregnancy Sickness and Meat
by Daniel M. T. Fessler
Pregnancy sickness, a suite of “symptoms” that frequently co-oc-
cur during pregnancy, may be an adaptation providing behavioral
prophylaxis against infection. Maternal immunosuppression, nec-
essary for tolerance of the fetus, results in gestational vulnerabil-
ity to pathogens. Throughout the period of maximal vulnerabil-
ity, dietary behavior is significantly altered via changes in nausea
susceptibility and olfaction and the development of marked aver-
sions and cravings. Of food types, meat is both the most likely
to carry pathogens and the principal target of gestational aver-
sions and pregnancy taboos. Because meat was prominent in an-
cestral human diets but hygienic procedures that effectively
eliminate the risk of meat-borne infection are recent, such path-
ogens likely constituted a source of selective pressure on preg-
nant females throughout human history. Both the relatively low
protein and energy demands of the first trimester and the exis-
tence of nonmeat alternatives would have allowed for the evolu-
tion of time-limited gestational meat-avoidance mechanisms.
Complementing these mechanisms, gestational cravings target
substances that may influence immune functioning and affect
the availability of iron in the gastro-intestinal tract, thereby limi-
ting the proliferation of iron-dependent pathogens. Clinical and
ethnographic findings are examined in light of these proposals,
and directions for future research are outlined.
daniel m. t. fessler is Assistant Professor of Anthropology
at the University of California, Los Angeles (Los Angeles, Calif.
90095-1553, U.S.A. []). Born in 1963,he
was educated at the University of California, San Diego (B.A.,
1985; M.A., 1988; Ph.D., 1995). He has conducted ethnographic
research in Indonesia, including a 32-month study of a fishing
village in Bengkulu, Sumatra. His publications include “Toward
an Understanding of the Universality of Second-Order Emo-
tions,” in Beyond Nature or Nurture, edited by A. Hinton (New
York: Cambridge University Press, 1999), “Emotions and Cost/
Benefit Assessment: The Role of Shame and Self-Esteem in Risk
Taking,” in Bounded Rationality, edited by R. Selten and G. Gig-
erenzer (Cambridge: MIT Press, 2000), and “Pseudoparadoxical
Impulsivity in Restrictive Anorexia Nervosa: A Consequence of
the Logic of Scarcity” (International Journal of Eating Disorders,
in press). The present paper was submitted 17 viii 00 and ac-
cepted 20 vi 01.
1. Paul Sherman and Samuel Flaxman graciously shared their pre-
publication results with me. C. David Navarrete wrote the resam-
pling program and performed the statistical analysis. The following
Around the world, people recognize that pregnancy is
associated with nausea, vomiting, and marked changes
in dietary preferences, a pattern termed pregnancy sick-
ness (Flaxman and Sherman 2000).
Nausea and vomiting
generally serve to protect the individual from the harm-
ful effects of ingested toxins and toxin-producing path-
ogens; vomiting expels the offensive substance, while
nausea, a highly aversive experience, anchors remarkable
one-trial learning, motivating future avoidance of the
substance (Bernstein 1999). The complex and coordi-
nated nature of these features, the specificity of their
eliciting conditions, and the obvious benefits that they
provide all suggest that normal nausea and vomiting are
part of an adaptation, a trait that has evolved through
natural selection because of the advantages that it fur-
nishes to its possessors. However, in contrast to the usual
circumstances associated with nausea and vomiting,
many pregnant women experience these symptoms after
ingesting foods which others around them eat without
ill effect—indeed, it seems that pregnant women often
become nauseated before ingestion has occurred and,
most notably, experience nausea and vomiting at far
higher frequencies than do other members of their com-
munities. The striking differences between the circum-
stances of normal nausea and vomiting and those of preg-
nancy (or “morning”) sickness call for explanation.
A large body of research addresses possible physiolog-
ical mechanisms underlying pregnancy sickness. Recent
individuals provided helpful comments on a draft of this paper: Paul
Sherman, Jonathan Haidt, David Haig, Susan Perry, Jim Moore, Rob
Boyd, Joe Manson, Dustin Penn, Thomas McDade, Nicholas Blur-
ton Jones, and five very diligent anonymous referees. I also benefited
from conversations with Jennifer Fessler, Joan Silk, Leda Cosmides,
and Sarah Hrdy. Paul Arguello and Karen Hunt Ahmed provided
research assistance, and John Fessler and Andre Kramerov provided
translation assistance. Portions of this project were presented at
the 2000 meeting of the Human Behavior and Evolution Society.
2. The Bengkulu of southwestern Sumatra term pregnancy “the
good sickness,” a reference to the nausea, vomiting, and strange
aversions and cravings which appear early in gestation. For the
!Kung San of the Kalahari desert, these are diagnostic criteria for
pregnancy (Shostak 1981). Soranus of Ephesus, a 2d-century Greek
gynecologist, described the same pattern in association with early
pregnancy (Temkin 1991), and the Maasai (Spencer 1988) and Tiv
(Bohannan and Bohannan 1969) do likewise. Cravings, nausea, and
vomiting are hallmarks of early pregnancy among the Hokkien Tai-
wanese (Barnett 1971) and Tongans (Ostraff et al. 2000, Morton n.d.);
Culpepper’s 1651 Directory for Midwives notes the same (Fair-
weather 1968), and these symptoms also occur among the Saraguro
(Finerman 1985) and Hawaiians (Green and Beckwith 1924). Unique
gestational cravings are recognized by the Jivaro (Harner 1972) and
the Tahitians (Levy 1973). Present-day midwives from such diverse
countries as Sierra Leone, Chile, the Netherlands, Pakistan, Va-
nuatu, Iran, Jamaica, and Nicaragua agree that nausea and vomiting
occur in early pregnancy (O’Brien, Relyea, and Lidstone 1997), and
anthropologists report the same for highland Ecuador (Weigel et al.
2000) and Truk (Fischer 1998 [1963]). Clinical surveys report nausea
and vomiting in association with pregnancy in 75 groups in Western
Europe, North America, Japan, India, Kenya, Nigeria, Sri Lanka,
and Hong Kong (Flaxman and Sherman 2000), and ethnographers
record the same among indigenous groups in North and South
America, the Caribbean, Oceania, Australia, Asia, the Middle East,
and Africa (reviewed in Weigel 1985 and Flaxman and Sherman
20 F current anthropology Volume
, Number
, February
investigations focus on specific isoforms of human cho-
rionic gonadotrophin (hCG) (cf. Jordan et al. 1999; also
O’Connor et al. 1998), with attention being paid to the
influence of hCG on thyroid functioning (Tareen et al.
1995). While some investigators have employed evidence
regarding proximate mechanisms to argue that preg-
nancy sickness is merely a nonfunctional or even dys-
functional by-product of hormonal changes accompa-
nying pregnancy (see Weigel 1985 for review and Profet
1992 and Flaxman and Sherman 2000 for additional dis-
cussion), others have postulated that pregnancy sickness,
like normal nausea and vomiting, is a functional adap-
tation. Specifically, the greater sensitivity of the mother
has been explained as a means of compensating for the
special vulnerability of the embryo or fetus (hereafter
generically termed the “developing organism”).
To date, adaptationist explanations have taken a num-
ber of forms. Hook (1976, 1978, 1980) and Tierson, Olsen,
and Hook (1985) initially proposed that pregnancy sick-
ness protects the developing organism from a heteroge-
neous collection of toxins such as those thought to occur
in coffee. Weigel (1985) generalized this list to include a
range of teratogens and pathogens. Noting that preg-
nancy sickness is concentrated in the first trimester,
when the developing organism is most vulnerable to in-
sult, Profet (1988, 1992) expanded the emphasis on te-
ratogens, arguing that pregnancy sickness principally
protects the developing organism from secondary com-
pounds produced by plants as defensive agents. Profet
also noted that cooked meat, which contains carcino-
gens, may be teratogenic and therefore could constitute
a secondary target of prophylactic avoidance.
If pregnancy sickness is an adaptation, natural selec-
tion should have imbued it with discriminatory speci-
ficity; in order to preclude the costly avoidance of ben-
eficial substances, only those foods that posed a
significant and recurrent threat should have acquired
nauseogenic salience during pregnancy. To test adapta-
tionist explanations, I therefore began compiling studies
of gestational food aversions. It soon became apparent
that, in general, contrary to Profet’s central thesis, plant
foods were not a central target of such aversions. Rather,
in most studies, the principal targets were meat and
other animal products. Troubled by an absence of evi-
dence in support of Profet’s secondary assertion that
cooked meat may be teratogenic and recalling Weigel’s
observation regarding the hazards of infection during
pregnancy (and Profet’s own brief reference to the risks
posed by spoiled meat [1992:334]), I noted that meat is
a common source of infection and pregnancy sickness
coincides with a period of marked maternal immuno-
vulnerability, a time of danger for both the developing
organism and the mother. While developing the ensuing
hypothesis that a principal function of pregnancy sick-
ness is the provision of prophylactic protection from
meat-borne pathogens during reproductive immunosup-
pression, I learned from Paul Sherman that he and Sam-
uel Flaxman had just completed a comprehensive meta-
analysis of studies of pregnancy sickness and had reached
the same conclusion. Flaxman and Sherman’s subse-
quent publication (2000) also documents the incidence
of pregnancy sickness in a large number of societies and
summarizes patterns of gestational cravings.
This paper builds on and complements Flaxman and
Sherman’s important work in a number of ways. First,
seeking additional evidence of the unique salience of
meat during pregnancy, I survey the targets of pregnancy
taboos in 73 cultures. Second, because adaptations are
thought to arise in response to highly specific selection
pressures, I document that (a) the timing of nausea, vom-
iting, and aversions coincides with the period of maximal
immunovulnerability and (b) meat-borne pathogens pos-
sess attributes that make them particularly dangerous
during precisely this period. Third, because adaptations
arise only in response to recurrent selection pressures, I
present evidence that meat-borne pathogens have prob-
ably constituted a source of selective pressure through-
out much of human history. Fourth, because intrinsic
trade-offs are thought to constrain the efficiency with
which any single adaptive goal is pursued, I demonstrate
that the benefits of avoiding meat during early pregnancy
are likely to have generally exceeded the costs. Fifth,
because adaptations are best identified on the basis of
the complex nature of their designs, extending sugges-
tions made by Profet and others I introduce evidence for
the existence of an integrated suite of sensory and in-
gestive changes in early pregnancy. I then explore the
possible utility of gestational cravings, arguing that they
often target substances that affect the availability of iron
in the gastrointestinal tract, thereby limiting the prolif-
eration of iron-dependent pathogens. Finally, I address
problems with the meat-avoidance hypothesis, compare
it with other explanations of pregnancy sickness and re-
lated phenomena, and present testable predictions de-
rived from it.
Meat Avoidance during Pregnancy
Building on the work of Hook (1976, 1978, 1980) and
Tierson et al. (1985), Profet (1988, 1992) has argued that
pregnancy sickness involves the patterned development
of a large number of marked aversions to previously fa-
vored foods. The acquisition of aversions is closely tied
to the experience of nausea, whether as a result of or in
anticipation of ingestion (reviewed in Flaxman and Sher-
man 2000; also Dye, Jones, and Hill 1998). Increased nau-
sea and vomiting are the hallmark of pregnancy sickness,
and, correspondingly, Flaxman and Sherman’s meta-
analysis reveals a substantial increase in the number of
aversions among pregnant versus nonpregnant women
(2000:129). However, while the association of gestational
aversions with nausea fits the general pattern, gesta-
tional aversions exhibit unique properties: In Western
subjects, whereas food aversions acquired in response to
nausea and vomiting generally persist for years or dec-
ades (Mattes 1991), nearly all gestational aversions dis-
appear with parturition (Worthington-Roberts et al.
1989, Pope, Skinner, and Carruth 1992). The combina-
tion of the greatly increased occurrence of aversions dur-
fessler Reproductive Immunosuppression and Diet F 21
Fig. 1. Average number of food aversions and cravings of pregnant women, all trimesters (Flaxman and Sher-
man 2000: fig. 7). Aversions based on
studies including
women and cravings on
studies including
women (see table
, P !
; NS, no significant difference. Meat, meats, fish, poultry, and eggs; N-A,
nonalcoholic beverages; Veg, vegetables; Alc, alcoholic beverages; ESS, ethnic, strong and spicy foods; D, dairy
products and ice cream; S, sweets, desserts, and chocolate; G&S, grains and starches; F, fruits and fruit juices.
ing pregnancy and their uniquely time-limited nature
supports Profet’s and others’ assertions that pregnancy
exercises a distinctive influence on food avoidance. In
order to determine whether gestational aversions are the
product of an evolved adaptation and, if so, what its func-
tion is, we must therefore establish which foods are
avoided during pregnancy.
Flaxman and Sherman (2000) conducted a meta-anal-
ysis of 20 studies of gestational aversions (covering 5,432
women) and 21 studies of gestational cravings (6,239
women) (see fig. 1) and found that meat is clearly the
principal target of aversions (see also Knox, Kremer, and
Pearce 1995, Dye, Jones, and Hill 1998). Although the
studies of aversions surveyed by Flaxman and Sherman
include investigations in Saudi Arabia, Jamaica, Sudan,
and South Africa, 16 of the 20 employ Western subjects.
In general, claims regarding species-typical adaptations
produced by natural selection are best supported by data
drawn from a wide range of human societies. Accord-
ingly, it is important to examine meat eating during preg-
nancy in a broader range of non-Western subjects. In
many traditional societies, food taboos and related be-
liefs are an important determinant of gestational inges-
tive behavior. While taboos may sometimes constitute
wholly exogenous constraints, a number of observations
suggest that they may in part reflect women’s own ex-
periences. Accordingly, any patterning of the categories
of foods targeted by taboos is of potential interest.
A search of the Human Relations Area Files, other
ethnographies, and the nutrition literature revealed 73
societies for which investigators provide information on
gestational food taboos. In tabulating the findings (tables
1 and 2), I employed the following conventions: In some
instances, an entire category of food was proscribed,
while in other cases only specific types of food were
taboo; in the latter, a given society was counted once
regardless of how many items were proscribed (because
categorical proscriptions cannot be quantitatively com-
pared with specific proscriptions, the units being com-
pared were whole societies rather than taboos). In some
cases there was a lack of concordance either between
two ethnographers or within one ethnography; I labeled
these “type B” cases, in contrast to the simpler “type
A.” In type B cases, positive evidence was given priority
over negative; since this was done for all food types, it
did not inflate the count for any given category. Likewise,
when one source indicated categorical proscriptions
while another indicated specific ones, only the former
was recorded. In the hopes of mitigating Galton’s Prob-
lem, data are presented by geographical region showing
the number of contributing societies.
3. The total number of societies listed in the two tables exceeds
the number of actual societies because a society appears in both
tables if it has categorical proscriptions for one food type and spe-
cific proscriptions for another food type.
22 F current anthropology Volume
, Number
, February
table 1
Targets of Taboos on General Categories of Food
Region and Type of
Report Number of Societies Meat Vegetables Fruit Dairy Products Sweets Spicy Foods Starches
North Africa
Type A
1 –– 1 ––
Type B
Sub-Saharan Africa
Type A
Type B
Europe and Euro-
Type A
1 ––1 ––
Type B
Middle East
Type A
1 –– 1
Type B
East Asia
Type A
2 1 –– 1
Type B
South Asia
Type A
1 –– 1
Type B
Southeast Asia
Type A
1 –– 1
Type B
Type A
Type B
Type A
Type B
Central America
Type A
Type B
11–– 1
South America
Type A
531 111
Type B
North America
Type A
21–– 2
Type B
Total 25 13 3 2 2 3 8 0
note: Type A, concordance among ethnographies; B, lack of concordance among or within ethnographies.
sources: 1, Calame-Griaule and Dieterlen (2000[1960]), Paulme (1940), van Beek (2000[1992]), Bohannan and Bohannan (1969); 2,
Kaufman (2000[1960]), Hanks (2000[1963]); 3, Kagwa et al. (1934), Mair (1934), Gilks and Orr (1931), Merker (1910), Spencer (1998),
Aunger et al. (1999), Israel and Nestor (1982a), Demissie, Muroki, and Kogi-Makau (1988), Ominde (1952); 4, Laguerre (1984); 5, Ab-
dalla (1982); 6, Pritham and Sammons (1993),Liulan (1999); 7, Stevenson (1920), Briggs (1920); 8, Manderson (1981); 9, Kaufman
(2000[1960]), Hanks (2000[1963]); 10,Ro´ heim (1993); 11, Whitehead (2000); 12, Chapin (1983), Nordenskio¨ld (1998[1938]); 13, Karsten
(1932), Peirson (1998[1967]), Elick (1969); 14, Turner (1972), Clark (1959).
Using these data, I conducted exact binomial tests of
the proposition that the proportion of meat taboos (de-
scribed in resampling terminology as “successes”) oc-
curring in this sample was equal to the proportion of
every other type of food taboo observed. The total num-
ber of successes was calibrated to reflect the proportion
of samples within each geographic region displaying the
taboo (i.e., 9 out of 10 societies within the same region
would return a success rate of 0.9 for that region). Type
A and type B ethnographies were pooled, and the number
of successes was rounded to the nearest integer for the
analysis. For the general food taboos, five out of six bi-
nomial tests executed returned p values less than 0.05
(see appendix). The only other food taboo that achieved
proportion levels statistically indistinguishable from
meat was spicy foods (cf. Flaxman and Sherman 2000).
When tested against the mean of all nonmeat taboos, the
p values were significant. For the specific food taboos all
seven tests returned p values less than 0.05. When the
data from general and specific food taboos were pooled
(taking care to ensure that only one type was counted),
all p values were below 0.05.Of20 binomial tests, 19
fessler Reproductive Immunosuppression and Diet F 23
Targets of Taboos on Specific Food Items
Region and Type of
Report Number of Societies Meat Vegetables Fruit Dairy Products Sweets Spicy Foods Starches
North Africa
Type A
Type B
1 –– 1 ––
Sub-Saharan Africa
Type A
17 13 1 5 2 2 2 6
Type B
Europe and Euro-
Type A
33–– 1 ––
Type B
Middle East
Type A
1 11 ––
Type B
East Asia
Type A
53 2 ––11
Type B
South Asia
Type A
Type B
Southeast Asia
Type A
5514 22 2
Type B
Type A
Type B
Type A
Type B
11 1 ––
Central America
Type A
Type B
South America
Type A
10 9 3 11
Type B
North America
Type A
33 1 ––
Type B
11 1 ––
Total 62 52 6 25 3 7 6 9
note: Type A, concordance among ethnographies; B, lack of concordance among or within ethnographies.
sources: 1, Calame-Griaule and Dieterlen (2000[1960]); 2, Kaufman (2000[1960]); 3, Brock (1953), Calonne-Beaufaict (1998[1921]),
Evans-Pritchard (1998[1932, 1937]), Richards (1939), Helitzer-Allen, Kendal, and Wirima (1997), Israel and Nestor (1981a, b, 1982b, c),
Costello (1986), Ominde (1952); 4, Pavlovic (1973), Sanders (1949), Fife and Fife (1966); 5, Abdalla (1982); 6, Pritham and Sammons
(1993), Wheeler and Tan (1983), Burkhardt (1954), Gould-Martin (1976), Liulan (1999); 7, De Silva (1995), Kar (1993), Ferro-Luzzi
(1973), Brown, Shah, and Worth (1968), Rizvi (1979), Panikkar (1918); 8, Adriani and Kruyt (1996[1951]), Man (1975), Bolton (1972),
Covarrubias (1938); 9,Ro´heim (1993); 10, Thompson (1940), Beaglehole and Beaglehole (1938, 1941), Jelliffe and Jelliffe (1964), Black-
wood (1935), Schieffelin (1990); 11, Malinowski (1929), Seligmann (1910); 12, Cosminsky (1994); 13, Chapin (1983); 14, Da Silva (1962),
Chapman (1982), Jackson (1983), Reichel-Dolmatoff (1971), Becher (1960), Taylor (1950), Parsons (1945), Bianco et al. (1990), Pereira et
al. (1989), Kracke (1981), Johnson (n.d.), Elick (1969); 15, Hilger (1951), Clark (1959); 16, Jones (1914), Emmons (1991).
returned p values less than 0.05, thus rejecting the null
hypothesis. This analysis affirms the profound overre-
presentation of meat taboos in the sampled societies.
Food taboos in pregnancy may reflect (a) the morali-
zation of a prevailing pattern (i.e., pregnant woman avoid
meat, and people notice this and create a rule), (b)an
institutionalized understanding of a causal relationship
(i.e., meat eating is observed to have negative conse-
quences and so is proscribed), or (c) a self-serving insti-
tution created for the benefit of those in power (i.e., meat
is valuable and men seek to monopolize it; pregnancy
epitomizes womanhood and therefore foregrounds rules
which benefit men at women’s expense). In addition,
meat may have special salience as a target of taboos, and
therefore pregnancy taboos may reflect a larger pattern
of proscriptions.
In some cases, the ethnographic record suggests pro-
cesses of introspection and observation. For example, Ar-
anda women describe numerous dreams linking preg-
nancy, bitter tastes, meat eating, and nausea. Some
24 F current anthropology Volume
, Number
, February
informants report that ”the woman may eat meat if she
wishes to do so, but the unborn child does not like it,
and may show its resentment by causing sickness”; oth-
ers state outright that meat should be avoided (Ro´heim
1933:251). In Tamilnad, India, meat and eggs are some-
times proscribed because they induce vomiting and also
because they are “hot” foods (Ferro-Luzzi 1973); “hot”
foods are often taboo during the first trimester because
they are thought to be abortifacient (Pool 1986). Papaya,
another “hot” food, is an even more prevalent target:
mango, elsewhere a common focus of cravings, is also
avoided because it is “hot” (Ferro-Luzzi 1973). While the
latex present in papaya skin does indeed induce uterine
contractions (Cherian 2000), there are no reports of such
an effect for mango. Therefore Tamilnad pregnancy be-
liefs suggest taboo formation based on observations of
causality combined with symbolic elaboration and ex-
Although the level of detail presented in most
ethnographies is insufficient to allow for similar analyses
of other cases, these examples suggest that gestational
proscriptions may be in part a function of women’s ex-
periences with meat.
In at least 10 of the 65 meat-proscribing cultures (Mat-
sigenka, Kaluki, Orang Asli, Island Carib, Eastern Toraja,
Andaman Islanders, Lau Fijians, Chippewa, Buka, and
Kagwahiv), patriarchy cannot explain the taboos, since
some or all meat is proscribed for both the pregnant
woman and her husband. Further evidence against pa-
triarchy as a universal explanation comes from the ob-
servation that these taboos often do not apply to pre- and
postreproductive-age women (Simoons 1994).
Additional ethnological analysis (reported elsewhere
[Fessler and Navarrete n.d.]) reveals that the patterned
focus on meat evident in pregnancy taboos also appears
in other food taboos, suggesting that similar factors un-
derlie the formation of both pregnancy and nonpreg-
nancy taboos. One possibility is therefore that pregnancy
taboos are not informative with regard to postulated en-
dogenous changes in women’s attitudes toward meat
during pregnancy. However, a second possibility is that
both pregnancy and nonpregnancy taboos focus on meat
because, given its unique combination of high nutri-
tional value and high risk of pathogen ingestion, humans
possess an evolved ambivalence toward it. Animal prod-
ucts have unparalleled salience as elicitors of disgust,
and in both humans and nonhumans proteinaceous foods
are the principal target of conditioned aversions. Food
taboos in general may reflect such reactions. Pregnancy
thus appears to entail an amplification of reactions that
are present outside of pregnancy, and hence pregnancy
taboos may spring in part from endogenous responses to
4. A common rationale is the (probably accurate) belief that meat
consumption increases birth weight, an effect which women seek
to avoid (despite the costs to infant survivorship) where labor is
seen as difficult and dangerous.
In sum, while it is impossible to discern the exact
causes of all gestational taboos, it is at least plausible
that the prominence of meat as a target of proscriptions
in part reflects women’s spontaneous subjective expe-
riences. Accordingly, a parsimonious explanation of the
concordance between Flaxman and Sherman’s and my
results is that the spontaneous avoidance of meat during
pregnancy is widespread across populations.
Timing of Pregnancy Sickness
As noted earlier, acquired food aversions are intimately
linked to nausea and vomiting. Numerous investigators
report that, in Western subjects, nausea and vomiting
are concentrated during the first trimester (Jarnfelt-Sam-
sioe, Samsioe, and Velinder 1983, MacIntyre 1983, Kle-
banoff et al. 1985, Weigel 1985, Tierson, Olsen, and Hook
1986, Vellacott, Cooke, and James 1988, Gulick, Shaw,
and Allison 1989, Rodin and Radke-Sharpe 1991, Gadsby,
Barnie-Adshead, and Jagger 1997; but see also Lacroix,
Eason, and Melzack 2000). The same pattern is reported
in widely disparate preindustrial societies. Consistent
with the link between nausea, vomiting, and acquired
aversions, Flaxman and Sherman (2000) demonstrate
that, in a subsample of the above studies of Western
subjects, pregnancy-related food aversions are signifi-
cantly concentrated in the first trimester and decline
thereafter (see also Knox, Kremer, and Pearce 1995).
The timing of pregnancy sickness has been singled out
by some adaptationists as evidence of evolved design. To
date, these arguments have focused on the delicate na-
ture of organogenesis, a process that occurs primarily
during the first trimester (Profet 1988, 1992; Flaxman and
Sherman 2000). Organ construction is equivalent to
shooting an arrow at a target—the consequences of a
slight deflection of the trajectory will be much greater
if they occur early in the arrow’s flight than if they occur
later. Similarly, early deviations in the organization of
organogenic cells have greater consequences than later
deviations. The period of organogenesis thus constitutes
a window during which insult to the developing organ-
ism, be it from teratogens or from infection, is likely to
have the most wide-ranging consequences (reviewed in
Flaxman and Sherman 2000). Moreover, the dangers in-
herent in this somewhat precarious phase are greatly ex-
acerbated by the fact that, during precisely this period,
the mother’s ability to resist infection is severely com-
5. Meat is so prominent in nonpregnancy taboos that even dramatic
increases in endogenous responses during pregnancy would not in-
crease the centrality of meat in pregnancy taboos relative to other
taboos. No claim is being made as to any functional benefit provided
by either pregnancy or nonpregnancy taboos. On the contrary, both
are best viewed as accidental consequences of the effects of indi-
vidual propensities on the generation and diffusion of cultural
knowledge (see Fessler and Navarrete n.d.).
fessler Reproductive Immunosuppression and Diet F 25
Reproductive Immunosuppression
Far from being a reflection of godly perfection, human
beings, like other living things, are largely a collection
of compromises. Given the precarious nature of embry-
onic development, in an ideal world women would be
most immunologically robust during early pregnancy.
However, this cannot occur because pregnancy brings
two dichotomous objectives into conflict. On the one
hand, pregnancy necessitates harboring a foreign body,
since half of the genetic material of the developing or-
ganism is paternally derived. On the other hand, the im-
mune system identifies and attacks foreign bodies. Ac-
cordingly, pregnancy can occur only if maternal
immunological defenses are initially restrained (see Loke
and King 1997).
Lymphocytes, crucial components of the immune sys-
tem, compose both specific and nonspecific defenses. T
and B cell lymphocytes are specific; each clonal cell line
binds only to a single antigen. Other lymphocytes, in-
cluding natural killer (NK) cells, are nonspecific; they
will attack any entity that shows indications of foreign-
ness or dysfunction. Lymphocytes both produce and are
in part regulated by a class of soluble polypeptides called
cytokines. Although the dynamics of cytokine-mediated
interactions among lymphocytes are complex, in general
there is an opposition between two types of cytokines,
referred to as Th1 and Th2.Th1 cytokines are generally
proinflammatory, that is, they promote aggressive cel-
lular response, notably including the excitation of NK
cells. Th2 cytokines are generally anti-inflammatory,
blocking the effect of Th1 cytokines and reducing the
responsiveness of NK cells (Szekeres-Bartho and Weg-
mann 1996).
In normal menstrual cycling, progesterone levels re-
main low through the follicular phase, begin to rise just
after ovulation, peak in the midluteal phase, and then
decline again by the end of the cycle. However, in the
event that conception occurs, instead of dropping, pro-
gesterone levels continue to climb, peaking in the first
trimester (Costea et al. 2000). High progesterone levels
stimulate the spleen to produce progesterone-induced
blocking factor (PIBF). PIBF stimulates production of Th2
cytokines (Szereday, Varga, and Szekeres-Bartho 1997),
with the result that there is a profound decrease in NK
cell activity (Szekeres-Bartho and Wegmann 1996). NK
cells, with their ability to attack any and all invaders,
constitute a fundamental threat to the developing or-
ganism, and hence progesterone (indirectly) reduces their
Progesterone also exercises direct effects, in-
hibiting lymphocyte proliferation in response to invasion
(Borel et al. 1999). Trophoblasts also produce progester-
one, increasing local concentrations of the hormone (see
6. Maternal immunotolerance is multifaceted. In the latter part of
the luteal phase CD56bright, a distinct type of NK cell, partially
displaces other NK cells. CD56bright is inhibited by trophoblasts
at the feto-maternal interface (Biassoni et al. 1999). The developing
organism is protected against T cells through induction of apoptosis
in cells which are sensitized to it, leading to the deletion of those
clonal lines (see Zhang et al. 1999).
Stites and Siiteri 1983 for review). As pregnancy pro-
gresses, progesterone production shifts from the corpus
luteum to the placenta (Hansen 1998), and there is a
corresponding change in the distribution of the hormone
over time, resulting in higher concentrations in the vi-
cinity of the developing organism. There is also a pro-
gressive rise in the production of anti-inflammatory cy-
tokines by cytotrophoblasts themselves, directly
inhibiting local NK activity (see Roth et al. 1996; Szer-
eday, Varga, and Szekeres-Bartho 1997). Together, this
multiplex restraint of lymphocyte activity allows for the
implantation of the (half-) foreign blastocyst in the en-
dometrium and the resulting development of the embryo
and placenta.
Although progesterone concentrations are highest at
the feto-maternal interface, systematically elevated pro-
gesterone levels are sufficient to suppress the immu-
noreactivity of NK cells throughout the body (see Szek-
eres-Bartho 1990). As a consequence, the pregnant
woman is vulnerable to infection by pathogens. More-
over, should infection occur, the feto-maternal interface
is especially susceptible to attack.
However, this weak-
ness is partially compensated for through an increase in
phagocytosis, the engulfing of diseased cells.
satory increases in phagocytosis take place progressively,
with the greatest increases in some forms of phagocytic
activity not occurring until the second trimester (Barriga,
Rodriguez, and Ortega 1994), while peak activity is
reached during the third trimester (Shibuya et al. 1987).
Likewise, although amniotic fluid stimulates the migra-
tion of leukocytes, presumably enhancing phagocytic de-
fense of the developing organism, it does not have this
effect until the second trimester (Gleicher et al. 1980).
Finally, it appears that some reduction in maternal im-
munosuppression occurs once gestation is well-estab-
lished, as progesterone levels peak in the first trimester
and there are progressive increases in maternal system-
atic levels of a number of Th1 cytokines following the
end of this period (Vassiliadis et al. 1998).
these patterns indicate that the initial phase of preg-
nancy constitutes a period of maximal maternal suscep-
tibility to pathogens.
Pathogens pose a threat not only to the mother but to
the developing organism as well. Eventually, the latter
will be protected by a second line of defense, as the pla-
centa is capable of generating nonspecific immune re-
actions (Paradowska et al. 1996). However, the true pla-
7. Systemic Th2 cytokine production increases progressively across
the three trimesters. Some investigators have found this to be
matched by increases in Th1 cytokine production, resulting in little
net change over pregnancy (Matthiesen et al. 1998), but others have
found the opposite pattern (Marzi et al. 1996).
8. Invasive tissues appear to be partially protected against maternal
macrophages (cf. Lu et al. 1989), possibly through the expression
of a special identifier protein (integrin-associated protein/CD47)(cf.
Oldenborg et al. 2000). Increases in systemic levels of phagocytosis
may be shackled to the maturation of the placenta because of the
importance of expression of CD47 by placental cells, thus sheltering
the developing organism from maternal macrophages.
9. An exception is a drop in IL-2 in the third trimester (Vassiliadis
et al. 1998).
26 F current anthropology Volume
, Number
, February
centa does not emerge until the 12th week, becoming
fully formed at 18 to 20 weeks (Beers and Berkow 1999).
Correspondingly, transfer of maternal gamma globulin
(IgG) greatly increases after the end of the first trimester,
providing the fetus with additional, specific, defenses
(see Webster 1998). The fetal immune system can ulti-
mately provide some defense against infection, but it is
likely that significant quantities of many of the most
important immune cells are not present until after the
first trimester (Webster 1998, Reha´kova´etal.1998, Sen-
ogles et al. 1979). Hence, for a number of reasons, the
vulnerability of the developing organism to pathogens is
likely to be highest at precisely the same time that ma-
ternal vulnerability is highest, in the first trimester.
The immunological vulnerability of the mother and
the developing organism during the first trimester is
likely to have constituted a source of selective pressure
during human evolution. Given that selection shapes not
only physiology but also behavior, one can ask whether
behavioral patterns can compensate for reduced patho-
gen resistance by decreasing the likelihood of transmis-
sion. In the case of some dangerous diseases, the answer
is clearly no. Cytomegalovirus, for example, although
not a threat to immunocompetent individuals, poses a
grave risk to pregnant women and their offspring (see
Forbes 1989), particularly early in pregnancy (Kumar and
Prokay 1983), but natural selection has not shaped fe-
male behavior so as to mitigate this danger. Transmis-
sion occurs through contact with infected individuals or
their urine. Children under five excrete more pathogens
than others and hence pose a threat to unborn siblings
(Bello 1992). To reduce the risk of infection, a pregnant
woman would therefore need to avoid her young children
during the first trimester, and in ancestral environments
this would have endangered valuable existing offspring.
In contrast to the adaptive dilemma posed by socially
transmitted pathogens such as cytomegalovirus, a sig-
nificant class of diseases that threaten pregnant women
can be avoided through affordable behavioral changes.
Ingestion is a common pathway for infection, and hence
increased dietary selectivity can have prophylactic ef-
Meat, Pathogens, and Pregnancy
Of all food types, animal protein (including meat, poul-
try, eggs, and seafood, hereafter generically termed
“meat”) is the most dangerous. Meat is the source of a
wide range of pathogens that pose a grave threat to preg-
nant women and developing organisms (see Smith 1999)
(table 3). Defense against these pathogens hinges upon
proinflammatory cytokine activity and, conversely, is of-
ten hindered by anti-inflammatory cytokines. Consis-
tent with the portrait of feto-maternal immunology pre-
sented earlier and in keeping with the susceptibility of
organogenesis to insult, these pathogens generally pose
the greatest risk early in pregnancy.
Many of the organisms listed in table 3 are present in
a wide variety of game animals, either pathogenically or
endosymbiotically. In some cases, pathogens are distrib-
uted throughout an infected animal; in other cases, they
are confined to an animal’s gastrointestinal tract but are
easily spread during butchering. Still others are simply
ubiquitous in the environment, with a propensity to pro-
liferate on meat. As a result of these factors, culturally
prescribed avoidance of particular species or selective
consumption of body parts provides little protection
from infection—all parts of any animal are potentially
dangerous. Although infected or contaminated meat can
be safely eaten if it is cooked thoroughly, this is not an
easy thing to accomplish, or gauge, when cooking over
a fire. Moreover, thorough cooking provides imperfect
protection if processing practices leave hands or imple-
ments contaminated. Although simple hygienic mea-
sures can prevent infection, such practices are likely to
be (and to have been) rare in most small-scale societies.
Accordingly, while the advent of cooking in hominid
evolution likely greatly reduced pathogen exposure as-
sociated with meat eating, for immunosuppressed fe-
males the residual risks will have remained significant.
Faunal remains and morphological features indicate
that meat has constituted a significant part of the diet
of our species since its inception, and it is likely that
this association dates to the very beginnings of our genus
(see Aiello and Wheeler 1995, Milton 1999, Sponheimer
and Lee-Thorp 1999). Although portions of some socie-
ties today are vegetarian, all available evidence indicates
that meat is likely to have played a critical role in the
diet of every member of our species prior to the advent
of agriculture (cf. Hill 1982, Kaplan et al. 2000, Mann
2000), and, while agriculture may have diminished the
importance of meat, it has not eliminated it. Together,
these factors make it likely that meat-borne diseases
have constituted a significant source of selective pres-
sure on pregnant women for much of human history.
If meat has been an important component in the hu-
man diet for most of our history, is it feasible (or, more
specifically, was it feasible in ancestral environments)
for pregnant women to avoid meat? As the case of cy-
tomegalovirus illustrates, if meat avoidance was prohib-
itively expensive, no amount of benefit gained would
have selected for the behavior. It is therefore important
that, from an energetic perspective, human pregnancy is
an extremely gradual affair: When scaled for maternal
body size, the daily increase in the energy demands of
pregnancy is lower in humans that in any other mammal
(Prentice and Whitehead 1987). The energetic require-
ments of the first trimester are low relative to those of
the second and third as a consequence of both the small
mass of the developing organism and associated tissues
and metabolic changes that may occur early in preg-
nancy, allowing more efficient use of available energy
(see Catalano 1999). Provided that sufficient energy is
consumed, dietary restriction during this period is there-
fore far less problematic than in later phases of preg-
fessler Reproductive Immunosuppression and Diet F 27
Attributes of Some Meat-Borne Pathogens That Pose a Threat to Pregnant Women
Distribution Sequelae
Limiting Cyto-
kine Response
Exacerbating Cyto-
kine Response
Period of Most Dan-
gerous Infection
Toxoplasma gondii All warm-blooded an-
imals can host;
common in many
mammals and
Abortion, stillbirth,
congenital deformi-
ties, fetal and ma-
ternal neurological
and retinal damage
Early pregnancy
Ubiquitous; infects or
is endosymbiotic
with many mam-
mals and marine
organisms; prolifer-
ates on meat
Maternal death, abor-
tion, neonatal
death, and menin-
1 25% fatality
rate with antibiotic
IL-8, GM-
Early pregnancy
Escherichia coli Ubiquitous; infects
many animals; pro-
liferates on meat
Maternal death, in-
trauterine death
TH2: IL-10
Early pregnancy
Brucella abortus, B.
Carried by many
mammals and ma-
rine organisms
Abortion, fetal death,
preterm delivery
Th2:IL-10, IL-4
Shigella dysenteriae Ubiquitous; often wa-
ter-borne, but pro-
liferates on meat
Maternal dehydra-
tion, electrolyte
imbalance, death,
abortion, preterm
Early pregnancy
Campylobacter jejuni Ubiquitous; infects or
is endosymbiotic
with many mam-
mals and birds; pro-
liferates on meat
Maternal dehydra-
tion, electrolyte
imbalance, death,
abortion, preterm
delivery, stillbirth
Th1: TNF-al-
pha, IL-8
Early pregnancy
Carried by many
mammals and
birds; proliferates
on meat
Leptospira Infects many mam-
mals and birds
Abortion, stillbirth
1 50% of cases
Th1: TNF-al-
pha, IL-2
Early pregnancy
sources: 1, Hejlı´cek, Litera´k, and Nezval (1997; 2, Remington, McLeod, and Desmonts (1995), Galva´n Ramı´rez et al. (1995); 3,
Thouvenin et al. (1997), Suzuki (1999); 4, Beazley and Egerman (1998; 5, Cooper and Walker (1998); 6, Farber and Peterkin (1991); 7,
kane et al. (1999); 8, Abram and Doric (1997); 9, Pepin, Russo, and Pardon (1997); 10, Moyo et al. (1995); 11, Chopra et al. (1997); 12,
Zimmer et al. (1996); 13, Sechser et al. (1976); 14, Corbel (1997); 15, Makhseed et al. (1998); 16, Oliveira et al. (1998); 17, Fernandes et
al. (1996); 18, Mwenye et al. (1997); 19, Sansonetti et al. (1999); 20, Cabrita et al. (1992); 21, Goh and Flynn (1992); 22, Pancorbo et al.
(1999); 23, Simor and Ferro (1990); 24, Lantelme et al. (1995); 25, Wallace et al. (1999); 26, Treml and Nesnalova (1993); 27, Carles et
al. (1995); 28, Petros, Leonhardt, and Engelmann (2000); 29, Shaked et al. (1993).
Moreover, meat is often of dietary importance
not as a source of calories but as a source of protein, and,
given the relative leanness of many game animals, this
10. Zhou, O’Brien, and Relyea (1999) found a negative association
between intensity of vomiting during pregnancy and birth weight
and consider it to be principally due to the link between the in-
tensity of vomiting and its duration across gestation. Women who
experience severe vomiting late in pregnancy are thought to suffer
inadequate nutritional intake during the period of maximal fetal
demands. Singh and Rosen (n.d.) found an association between adult
fluctuating asymmetry and maternal third- (but not first- or sec-
ond-) trimester nausea, a finding which can be interpreted as in-
dicating that the negative impact of pregnancy sickness–induced
dietary restriction is greatest late in pregnancy. These findings can
be viewed as suggesting that the benefit/cost ratio of pregnancy
sickness decreases across gestation, and this is consistent with its
predominance early in pregnancy.
is particularly likely to have been true in ancestral for-
aging populations (Milton 1999; also Mann 2000).
In animal models, protein restriction during early preg-
nancy stimulates placental growth (Perry et al. 1999). In
a study of a Dutch famine, Lumey (1998) shows that
human maternal undernutrition during the first trimes-
ter increases placental weight. Meat supplies are quickly
exhausted during famines, and hence Lumey’s results
suggest that humans share the pattern of enhanced pla-
cental growth in response to protein restriction. The
probable goal is a compensatory increase in the surface
area across which nutrients from the mother are supplied
to the developing organism (see Dwyer et al. 1995). There
is a ceiling to such benefits; a large ratio between the
weight of the placenta and that of the infant is predictive
28 F current anthropology Volume
, Number
, February
of future health problems (Langley-Evans et al. 1996).
However, such disproportionate ratios most likely occur
when the maternal diet is restricted throughout preg-
nancy, a condition that impacts fetal health in a variety
of ways. Although prolonged protein deprivation during
pregnancy is detrimental, decreased protein intake that
is restricted to the first trimester may be minimally
harmful (cf. Langley-Evans et al. 1996), harmless (cf.
Dwyer et al. 1995, Muaku et al. 1995), or even beneficial
(cf. Perry et al. 1999). Lastly, while meat is an important
source of protein, it is not the only source, and hence it
is possible that compensatory increases in the con-
sumption of other foods might mitigate the reduction in
protein intake entailed by meat avoidance.
Hypoferraemia (iron-deficiency anaemia) during the
first trimester is associated with poor pregnancy out-
come (see Schwartz and Thurnau 1995). Meat is a prin-
cipal source of iron, a critical determinant of cellular
growth, and hence avoidance of meat early in pregnancy
could conceivably have detrimental consequences. How-
ever, as I will show, there is evidence that mild hypo-
ferraemia may actually be beneficial and that iron avail-
ability is influenced by multiple dietary factors.
In sum, well-nourished women can apparently afford
to avoid meat during the first trimester, making it pos-
sible for selection to have shaped the psychophysiolog-
ical determinants of behavior in such a way as to shield
immunosuppressed pregnant women against the threat
of meat-borne pathogens.
Additional Design Features of Pregnancy
Having established that, as least with regard to the haz-
ards posed by meat-borne pathogens, pregnancy sickness
evinces the outward signs of an adaptation produced by
natural selection, keeping in mind Williams’s (1966) ad-
monition that adaptations are best identified on the basis
of the complex nature of their designs I turn now to a
more detailed examination of the subjective changes that
affect ingestion during pregnancy.
Profet (1992) and Erick (1995) suggest that the increase
in nausea susceptibility and changes in food preferences
co-occur with an increase in olfactory acuity (cf. Dickens
and Trethowan 1971). This is a plausible suggestion
given that olfaction is a principal avenue for the remote
detection of food properties and plays an important role
in the elicitation of nausea (cf. Ferna´ ndez-Marcos et al.
Yeremina (1976) examined 439 pregnant Siberian
women and found enhanced epithelial motility and ol-
factory sensitivity during the first and second trimesters,
but methods are not described. A longitudinal study of
20 German women failed to find an influence of either
pregnant state or stage of pregnancy on olfactory acuity.
However, compared with controls, pregnant women
rated four of five food-related odors as less palatable, a
trend which was most significant for peanut odor (Laska
et al. 1996). Gilbert and Wysocki (1991) compared 13,610
pregnant women with 277,228 nonpregnant women be-
tween the ages of 20 and 40 in the United States and
found that pregnant women failed to display enhanced
acuity to five of six odorants but (a) were less willing to
eat three of the six odorants, (b) rated three odorants as
less pleasant than they were rated by nonpregnant
women, and (c) were less willing to wear perfume.
Hansen and Glass (1936) report “olfactory perver-
sions” among pregnant German women: Individuals
who have spent their whole lives in animal husbandry
describe being unable to bear the smell of animals, par-
ticularly their excreta, when pregnant, and the same is
true with regard to fish odor.
Fairburn, Stein, and Jones
(1992:667) recorded gestational aversions among 80 out
of 100 subjects and noted that “an altered sense of smell
appeared to underlie the majority of the aversions.”
O’Brien, Relyea, and Lidstone (1997) asked 124 women
to record elicitors of nausea during pregnancy; odors were
highly nauseogenic, an effect which subjects blamed on
profound alterations in the sense of smell. Knox, Kremer,
and Pearce (1995) asked 100 pregnant Irish women about
alterations in smell and found that 39% reported height-
ened olfactory awareness; this change was strongly cor-
related with experiencing cravings and aversions (fried
food and cooking meat being particularly aversive) but
not with the incidence of nausea or vomiting. Three days
or less postpartum Cooksey (1995) asked 280 women,
“Did you experience craving to smell anything special?”
Fourteen percent reported strange olfactory cravings
never before experienced, and 1.4% spontaneously re-
ported alterations in olfaction. Cantoni et al. (1998) col-
lected retrospective data from 500 German women, find-
ing that 68% reported changes in olfactory perception
(principally increased olfactory responsiveness) during
pregnancy, most markedly during the first trimester. Sev-
enty-four percent remembered perceiving specific odors
as markedly less pleasant than usual, while 21% recalled
increases in olfactory pleasure; meat and smoke were
prominent stimuli for the former class, while fruit and
flowers were prominent for the latter. Fifty-eight percent
reported nausea in response to odors, and 75% reported
changes in diet, with both effects concentrated in the
first trimester. Changes in olfactory perception were pos-
itively correlated with both nausea and changes in
Looking beyond Western cultures, Al-Kanhal and Bani
(1995) list “unpleasant smell” as an important reason
given by pregnant Saudi women for avoiding particular
foods. Morton (n.d.) describes a powerful association be-
tween odor and nausea among pregnant Tongan women,
and Beaglehole and Beaglehole (1941) similarly report
both reduction in appetite and abnormal sensitivity to
bad smells as indices of the onset of pregnancy in Tonga.
Obeyesekere (1985) reports “olfactory distortions” in-
11. The researchers also tested olfactory acuity using the odors of
rubber, rose, and almonds. They describe a deficit in the third tri-
mester relative to the postnatal period but do not report the results
from their examinations earlier in pregnancy.
fessler Reproductive Immunosuppression and Diet F 29
volving revulsion among pregnant women in Sri Lanka.
Beaglehole and Beaglehole (2000 [1935]) note that early
in pregnancy the Hopi woman is “without appetite, since
food, especially meat, smells bad.” In conclusion, al-
though acuity per se may be unchanged in pregnancy,
marked alterations do indeed occur in olfactory
As Profet (1992:334) notes, olfactory cues are an ex-
tremely accurate index of degree of spoilage caused by
bacteria in meat. Although systematically collected data
on olfactory preferences are generally absent from the
ethnographic corpus, it is my impression that, while cul-
tural variation exists, such putrid odors are nauseogenic
in most populations (cf. Tuzin 1986)—selection seems
to have crafted an aversion to these odors, presumably
because of the risk of pathogen ingestion. Correspond-
ingly, mechanisms that increase the aversiveness of pu-
trid odors during pregnancy are likely to decrease the
probability that spoiled meat will be eaten. However, an
exclusive focus on putrid odors would provide only par-
tial protection, since many pathogens, particularly those
that are endosymbiotic with prey species, are unlikely
to produce a detectable odor. Accordingly, the goal of
prophylaxis would be best served by hedonic changes
such that the odor of meat in general is aversive. A plau-
sible phylogenetic scenario is therefore one in which pu-
trid meat was the initial target of olfactory safeguards
and these measures were subsequently expanded to in-
clude animal flesh in general (J. Moore, personal commu-
Profet (1992:334) suggests that the odor of cooking
meat is particularly aversive for pregnant women, and
Knox, Kremer, and Pearce (1995) robustly confirm this.
Cooking is likely to release many of the volatile con-
stituents of meat odor and can thus be seen as an am-
plification of the aversive signal. Flaxman and Sherman
(2000:127) report that, second to meat, a principal target
of gestational aversions is “nonalcoholic beverages,” of
which coffee is the primary contributor. As Profet (1992:
352) points out, odor is an important element in gesta-
tional aversions to coffee (cf. Knox, Kremer, and Pearce
1995). On the plant, however, coffee cherries have little
odor, and this suggests that, contrary to Profet’s phyto-
toxin hypothesis, the aversion to coffee may be a con-
sequence of the misfiring of a mechanism designed to
detect volatile meat odors.
Whereas odor seems more likely to signal danger in
meats than in vegetable foods, the converse would seem
to be true of taste. Arguing that bitter tastes index phy-
totoxicity, Profet (1992:333) postulated an increased
aversion to bitter tastes in pregnancy. This prediction
has recently been confirmed by Duffy et al. (1998), mak-
ing it more difficult to dismiss the phytotoxin hypothesis
despite the low frequency of reported aversions to
12. Sipiora et al. (2000) found that hyperemetic pregnant women
exhibited different patterns of bitter-taste perception from those
with lower levels of vomiting.
Gestational Cravings
Both the ethnographic and the clinical literature report
marked cravings in conjunction with pregnancy. A num-
ber of investigations have considered the possibility that
such cravings may have adaptive value; typically, expla-
nations focus on the ingestion of sugars or micronutri-
ents (cf. Jeans, Smith, and Stearns 1952, Hook 1980, Tier-
son, Olsen, and Hook 1985, Demissie, Muroki, and
Kogi-Makau 1998). Consistent with the changes in ol-
factory experience discussed above, there appears to be
an association between cravings and olfaction (Pelchat
and Schaefer 2000, Cooksey 1995), and cravings appear
primarily in the first trimester (Wijewardene, Fonseka,
and Goonaratne 1994, Flaxman and Sherman 2000). As
is true of aversions, cravings are clearly motivationally
relevant, as both experiences importantly influence
women’s diets (Harries and Hughes 1957; Hook 1978,
1980; Darwish, Amine, and Abdalla 1982; Schwab and
Axelson 1984; Tierson, Olsen, and Hook 1985; Knox,
Kremer, and Pearce 1995; Pope, Skinner, and Carruth
dairy products and protein
Flaxman and Sherman report a high frequency of crav-
ings for milk and other dairy products and a low fre-
quency of aversions. While the attractiveness of meat
declines during pregnancy, the attractiveness of other
sources of protein may rise in a complementary fash-
ion—the propensity to experience cravings is signifi-
cantly associated with the propensity to experience aver-
sions (Knox, Kremer, and Pearce 1995), suggesting a
compensatory mechanism (Demissie, Muroki, and Kogi-
Makau 1998). Such compensatory adjustments can ex-
plain the lack of negative impact of nausea and vomiting
(severity of which is associated with number of aversions
[Rodin 1989, Crystal, Bowen, and Bernstein 1999]) on
postpartum measures of maternal nutritional status
(Weigel et al. 2000).
Humans may be able to detect protein using olfactory,
gustatory, and possibly postingestive cues (cf. Wester-
terp-Plantenga and Wijckmans-Duijsens 1996, Crovetti
et al. 1998). By increasing the appeal of cues of protein
(while still making meat aversive), some of the costs of
meat avoidance could be reduced. In ancestral environ-
ments, plant foods, such as nuts and seeds, and insects
would have been the primary sources of nonmeat protein
(Southgate 1991).
Cravings for dairy products may thus
13. In addition to protein, important nutrients include the essential
fatty acids necessary for proper neurological development. As is
true of protein, many of these nutrients can be obtained from nuts
and seeds (Ollis, Meyer, and Howe 1999). In contemporary circum-
stances, nuts and seeds may sometimes be hazardous to pregnant
women because they contain mycotoxins produced by molds (Per-
aica et al. 1999). However, the risk of mold proliferation is greatest
when nuts or seeds are stored in large quantities (Peraica et al. 1999),
and hence it is not clear that mycotoxins would have constituted
a significant source of selective pressure prior to extensive reliance
on agriculture. Nevertheless, given that nuts and seeds may contain
30 F current anthropology Volume
, Number
, February
constitute an evolutionarily recent manifestation of a
mechanism designed to motivate the search for alter-
native protein sources.
clays and calcium
Geophagy is found in many societies (Reid 1992, Wiley
and Katz 1998, O’Brien, Relyea, and Lidstone 1997). De-
spite persistent efforts to classify it as a pathology or
culture-bound syndrome (cf. Grigsby et al. 1999), it is
likely that geophagy is often adaptive (see Reid 1992,
Wiley and Katz 1998). Animal models indicate that ge-
ophagy is often elicited by nausea (Mitchell, Winter, and
Morisaki 1977, McCaffrey 1985). Many primates and
other animals engage in geophagy in conjunction with
the consumption of substances that contain high con-
centrations of toxins (Krishnamani and Mahaney 2000,
Gilardi et al. 1999). Provenancial evidence suggests that
hominid geophagy may date back to Homo erectus,if
not beyond (Johns and Duquette 1991). Both nonhuman
animals and humans are generally quite selective about
the type of earth consumed, and olfactory cues may be
significant in the attraction to particular earths (Forsyth
and Benoit 1989, Krishnamani and Mahaney 2000;cf.
Harner 1972).
Clays, the most frequent targets of ge-
ophagy, possess a structure that makes them extremely
absorbent, resulting in remarkable detoxifying attributes
(see Johns and Duquette 1991, Reid 1992, Wiley and Katz
1998, Krishnamani and Mahaney 2000). Given its dis-
tribution across taxa, it is likely that there is an evolved
attraction to clays during nausea; societies may then rec-
ognize this attraction, as well as its beneficial conse-
quences, and institutionalize the behavior (Wiley and
Katz 1998).
Geophagy is associated with pregnancy, sometimes ex-
clusively so, in many societies (Wiley and Katz 1998; cf.
Harner 1972). Both the greater incidence of nausea and
the greater importance of detoxification are likely to re-
sult in increases in the behavior during this period (Wiley
secondary compounds (cf. Gilardi et al. 1999) (and given Laska et
al.’s olfactory data), broad generalizations about the costs and ben-
efits of nut and seed consumption in ancestral environments are
premature. Eusocial insects, the preferred targets of human con-
sumption (Defoliart 1995), themselves face strong pathogen risk
and may have evolved hygienic practices and other specialized ad-
aptations in response (cf. Rosengaus et al. 1998, 1999). As a con-
sequence, at least with regard to those pathogens that afflict both
humans and their insect prey, eusocial insects may be a less risky
source of protein than meat.
14. An obstacle to adaptive explanations of dairy-product cravings
is the fact that dairy products are a potential source of many of the
pathogens described above (Vasavada 1988). Conventional evolu-
tionary psychology (cf. Tooby and Cosmides 1992) holds that the
approximately 6,000 years (Bo¨ko¨nyi 1969) since humans are
thought to have begun consuming the milk of domesticated ani-
mals is too short a period for the evolution of an aversion to milk.
However, the presence of adult lactase production in populations
with a history of animal milk consumption (see Durham 1991)
suggests that adaptations may evolve or be modified within such
a period.
15. Geophagy carries the risk of infection from soil-borne parasites
(Geissler et al. 1997). However, children seem particularly likely
to suffer this cost, perhaps as a result of inadequate selectivity.
and Katz 1998; Profet 1992:345). However, although ge-
ophagy is present across Africa, it is rare or absent in
societies in which pregnant women consume significant
quantities of milk and other dairy products (Wiley and
Katz 1998). Milk differs from clays in that it lacks the
surface-area properties that make the latter highly ab-
sorbent. Correspondingly, milk is a relatively poor de-
toxifier and has no antinauseogenic properties (cf. Hu,
Lagomarsino, and Luo 1998, Lindseth and Lindseth
1995). However, both many of the targeted clays (Wiley
and Katz 1998) and milk possess a high calcium content.
Calcium demands rise during pregnancy as a result of
both the increase in maternal plasma and the developing
organism’s progressively escalating requirements. These
increased demands are anticipated by enhanced maternal
intestinal absorption of calcium beginning in the first
trimester (Ritchie et al. 1998). Both fetal demand and
maternal absorption increase markedly during the sec-
ond trimester, the period of fetal skeletal formation.
Grave consequences accompany a failure of the latter to
keep pace with the former.
Preeclampsia is a potentially dangerous condition
characterized by maternal hypertension after the 20th
week and a shift to proinflammatory cytokines and cel-
lular activity (Saito et al. 1999). Inadequate calcium con-
sumption increases the risk of preeclampsia, and sup-
plementation reduces it (Crowther et al. 1999,
Lopez-Jaramillo 2000), apparently because of the role
which calcium plays in the system whereby diseased
cells incite immunological response (cf. von Dadelszen,
Wilkins, and Redman 1999, Hojo et al. 1999). Demands
for calcium thus derive from both the structural impor-
tance of calcium for the developing organism and the
necessity of maintaining sufficient calcium levels to pre-
clude maternal immunological rejection of it. These con-
siderations are so important that maternal bone matter
is routinely sacrificed to ensure adequate calcium levels
(cf. Black et al. 2000). Increases in the subjective appeal
of calcium-rich clays and dairy products during the first
and second trimesters may thus be understood as an ad-
aptation designed to enhance the stockpiling and, later,
the supply of calcium in order to meet increasing fetal
demands for calcium (cf. Hook 1978, 1980; Wiley and
Katz 1998).
Iron is required for all cellular growth. However, un-
like calcium, iron is toxic at high concentrations, and
the body is unable to store substantial quantities for fu-
ture needs. Moreover, to make matters worse, the pres-
ence of calcium inhibits the absorption of dietary iron
and meat is a principal source of iron. Hence, when
viewed with regard to the developing organism’s need
for iron, the existence of both cravings for calcium and
aversions to meat appear bizarre and maladaptive. How-
ever, cravings for fruit shed light on the resolution to the
apparent paradox posed by the need simultaneously to
avoid pathogens, ensure an adequate calcium supply, and
maintain iron levels.
fessler Reproductive Immunosuppression and Diet F 31
fruit and vitamin c
Flaxman and Sherman (2000:127) find that fruit is the
single most common target of cravings and is rarely aver-
sive. For Bengkulu women, fruit is the prototypical food
craved during pregnancy, and, similar patterns appear
In ancestral environments, fruit is likely to
have been an important source of carbohydrates (South-
gate 1991). The human preference for sweet tastes, a
prominent factor in the desire for fruit, is most likely an
adaptation that facilitates caloric maximization. Sweet
foods in general are a prominent target of gestational
cravings (Flaxman and Sherman 2000, Taggart 1961, Ob-
eyesekere 1985, Wijewardene, Fonseka, and Goonaratne
1994, Knox, Kremer, and Pearce 1995). Hook (1978, 1980)
argues that these cravings facilitate the expansion of en-
ergetic reserves in anticipation of the demanding later
phases of pregnancy.
A number of sources report cravings for unripe fruit
(Adriani and Kruyt 1996 [1951], my field notes, Wije-
wardene, Fonseka, and Goonaratne 1994). Likewise, in
disparate populations many subjects experience cravings
for sour or savory foods.
In comparison with a simple
desire for sweet things, such preferences seem less likely
to maximize the caloric content of foods. Rather, the
craving for sweet and the craving for sour are related, as
fruits often combine these two tastes—in the ancestral
past, fruit was the principal source of these gustatory
For much of human history, fruit was a key source of
vitamin C (Southgate 1991). There is a clear correlation
between fruit consumption during pregnancy and serum
vitamin C levels (Ortega et al. 1998). Vitamin C provides
protection against a number of teratogenic substances
and unfavorable uterine conditions (cf. Usami, Tabata,
and Ohno 1999, Maritz and van Wyk 1997). Animal mod-
els and human studies suggest that vitamin C enhances
proinflammatory cytokine production (Schwager and
Schulze 1998, Jeng et al. 1996), thus partially counter-
acting the shift to anti-inflammatory cytokines engen-
dered by high progesterone levels. Vitamin C may also
greatly enhance the productivity of activated macro-
phages (Mizutani et al. 1998). While such actions might
seem to expose the developing organism to maternal im-
munological attack, this danger is perhaps mitigated by
vitamin C’s simultaneous reduction of NK cell activity
16. See Taggart (1961 [Scottish]), Obeyesekere (1985) and Wijewar-
dene, Fonseka, and Goonaratne (1994 [Sri Lankan]), Dye, Jones, and
Hill (1998 [English]), Thompson (1940 [Fijian]), Textor (1973 [Thai]),
Knox, Kremer, and Pearce (1995 [poor Irish]), cf. Norgan, Ferro-
Luzzi, and Durnin (1974 [New Guinean]), Dufour, Reina, and Spurr
(1999 [Colombian]).
17. See Harries and Hughes (1957), Fairburn, Stein, and Jones (1992
[English]), Taggart (1961 [Scottish]), Knox, Kremer, and Pearce (1994
[Irish]), Posner, McCottry, and Posner (1957), Schwab and Axelson
(1984), Finley et al. (1985), Tierson, Olsen, and Hook (1985), Pope,
Skinner, and Carruth (1992 [U.S.]), Kehrer (reported in Hansen and
Langer 1935 [German]), Al-Kanhal and Bani (1995 [Saudi]), Walker
et al. (1985 [South African]), Rizvi (1980 [Bangladesh]), Obeyesekere
(1985), Wijewardeme, Fonseka, and Goonaratne (1994 [Sri Lankan]),
Foll (1959 [Burmese]), Laderman (1981 [Malay]), Textor (1973
[Thai]), Kaufman (2000 [1960] [Thai]), Levy (1973:143 [Tahitian]).
(Huwyler, Hirt, and Morell 1985). Mathews, Yudkin, and
Neil (1999) found that (a) vitamin C was the only mi-
cronutrient predictive of birth weight and placental
weight (recall that a larger placenta is a good thing, pro-
vided that birth weight rises as well) and (b) these effects
occurred only with regard to vitamin C consumption
early in pregnancy.
A craving for foods high in vitamin
C during the first trimester thus ensures a supply of the
micronutrient which suffices to meet the needs of both
the mother and the developing organism.
Vitamin C enormously enhances the absorption of
iron, in part through preventing the formation of insol-
uble iron compounds. It thus counteracts the detrimen-
tal influence of calcium on iron absorption (see Lynch
and Cook 1980). One simple possibility, therefore, is
that, while meat avoidance reduces dietary iron intake
and calcium consumption reduces absorption, the re-
sulting decrement in iron availability is compensated for
through increased vitamin C consumption. Although the
basic outlines of this explanation may be correct, actual
events are likely more complex.
Gestational Hypoferraemia
I noted earlier that extreme hypoferraemia during the
first trimester can have detrimental consequences. How-
ever, pregnancy is frequency characterized by a progres-
sive increase in hypoferraemia to the point that, by the
third trimester, the vast majority of pregnant women
have an iron profile equivalent to that of an extremely
iron-deprived nonpregnant woman (Holla´n and Johansen
1993). Fetal demands for iron increase progressively and
do so independent of the level of maternal reserves (Bent-
ley 1985). Because iron levels are relative to total blood
volume, the significant increase in plasma also neces-
sarily results in lower iron levels (Holla´n and Johansen
1993). Finally, iron regulation may play a critical role in
fighting disease.
Because microbial and protozoan pathogens are de-
pendent on iron, the availability of iron can be a limiting
factor in pathogen proliferation. Accordingly, the body’s
response to pathogen invasion can be viewed as a delicate
balancing act between meeting the need for iron and
withholding iron from the invader (see Sunder-Plass-
mann, Patruta, and Horl 1999 for review). Hypoferraemia
in general may thus constitute an adaptive response to
pathogen threat (Holla´n and Johansen 1993, Nesse and
Williams 1995). Likewise, gestational hypoferraemia is
probably both normal and highly functional (Goodlin
1982, Montgomery 1990, Holla´n and Johansen 1993).
There is a strong association between maintenance of
high iron levels and the risk of low birth weight and
preterm delivery (Goldenberg et al. 1996), outcomes as-
sociated with maternal infection. All of the meat-borne
18. Mathews et al. play down the clinical significance of these
findings for well-nourished British women; however, these effects
are likely to be amplified under conditions more typical of ancestral
32 F current anthropology Volume
, Number
, February
pathogens listed in table 3 are highly dependent on iron
(Dimier and Bout 1998, Fisher and Martin 1999, Lopez-
Goni and Moriyon 1995, Chart, Scotland, and Rowe
1989, Coker and Obi 1991, but see also Schwartz et al.
1996, Miles and Maskell 1985, Staneck, Henneberry, and
Cox 1973).
The threat of disease is greatest during the first tri-
mester, yet hypoferraemia increases over the course of
pregnancy. If hypoferraemia provides disease protection,
why is the pattern not reversed? First, in keeping with
the importance of iron for cell growth and replication,
iron transfer occurs as soon as sufficient placentation has
taken place (Gulbis et al. 1994), and first-trimester hy-
poferraemia is associated with negative birth outcomes
(Schwartz and Thurnau 1995). Second, maternal iron lev-
els must not only accommodate immediate fetal de-
mands but also be calibrated in light of the upcoming
progressively increasing fetal needs. Maternal iron ab-
sorption rates increase ninefold from the first trimester
to the third (Barrett et al. 1994). Nevertheless, despite
increased absorption, maternal iron levels continue to
drop with each week of gestation (Milman, Graudal, and
Agger 1995)—a trend which is evidently due to increas-
ing fetal demands, since the effect is more substantial
in multiple gestations (Ben Miled et al. 1989). Therefore,
although hypoferraemia during the first trimester could
conceivably provide enhanced protection from patho-
gens, both the initial demands of the developing organ-
ism for iron and the rapidity of the increase in those
demands preclude beginning the developmental process
with low maternal iron levels. The converse considera-
tions explain why rates of maternal absorption do not
increase sooner. Although greatly enhancing iron ab-
sorption during the first trimester would decrease the
strain which the developing organism places on maternal
iron stores, an abundance of iron during the period of
maximum immunovulnerability would facilitate path-
ogen proliferation. The lag between increases in absorp-
tion rates and increases in demands on maternal stores
can thus be seen as an attempt to minimize disease vul-
nerability while still meeting the needs of the developing
Many ingested pathogens begin the process of prolif-
eration prior to extensive tissue penetration. Accord-
ingly, the presence of iron in the gastrointestinal envi-
ronment may constitute a critical determinant of the
hazards posed by ingested pathogens (Stockman 1981).
A number of primates consume calcium-rich clays when
suffering heavy parasite loads (Krishnamani and Maha-
ney 2000), one benefit of which may be a reduction in
the availability of iron in the gastrointestinal tract. Many
studies find an association between geophagy and ane-
mia in humans. Although some investigators, noting the
presence of iron in soils, have argued that anemia pre-
cedes and motivates geophagy, the preferential focus on
calcium-rich clays with a marked ability to inhibit the
absorption of iron suggests that causality occurs in the
opposite direction (see Reid 1992, Patterson and Staszak
1977, Hochstein 1968).
Geophagy is a form of pica, the ingestion of inedible
substances. Gestational pica occurs at differing frequen-
cies in diverse populations but is likely grossly under-
reported in some groups given the atmosphere of shame
and secrecy which often surrounds it (Cooksey 1995).
Pica provides definitive evidence of purely endogenous
modification of gestational ingestion, since overwhelm-
ing compulsions are experienced even by women who
are unaware that others have similar experiences, are
baffled by their desires, and worry that their behavior is
“crazy.” These cravings lead to ingestion in spite of fears
that it will harm the developing organism or the mother
herself. Olfaction appears to play a central role in pica
cravings (Cooksey 1995, Simpson et al. 2000).
Inedible substances ingested as a consequence of ges-
tational cravings include cornstarch, laundry starch,
plaster, ashes, coal, chalk, burnt matches, baking soda,
baking powder, flour, cigarette ashes, egg shells, mag-
nesium carbonate, powdered detergent, toothpaste,
whitewash, corn meal, raw oatmeal, raw rice, raw po-
tatoes, dry powdered milk, and baby powder (Jeans,
Smith, and Stearns 1952, Harries and Hughes 1957, Pos-
ner, McCottry, and Posner 1957, Walker et al. 1985,
Cooksey 1995, Al-Kanhal and Bani 1995, Knox, Kremer,
and Pearce 1995, Simpson et al. 2000, Smulian, Moti-
wala, and Sigman 1995, Rainville 1998, Rizvi 1980; also
Levy 1973; O’Brien, Relyea, and Lidstone 1997). Many
of these materials are highly absorbent, and, as in the
case of geophagy, this may allow them to mitigate the
danger posed by ingested toxins. Likewise, just as ge-
ophagy occurs among many animals, some nonhuman
primates consume charcoal in response to high levels of
phytotoxins (Krishnamani and Mahaney 2000). Both the
significant toxin-neutralizing properties of many clay
and nonclay targets of pica and the presence of compa-
rable pica in nonmeat-eating species lend support to Pro-
fet’s claim that phytotoxin avoidance is an important
adaptive function of pregnancy sickness.
Other forms of pica may also have multiple benefits.
In addition to being highly absorbent, many of the sub-
stances listed above have a marked ability to inhibit the
absorption of iron.
Some, such as chalk, plaster, white-
19. Clinical trials showing that iron supplementation reduces ge-
ophagy and related practices have been adduced as evidence that
anemia causes geophagy. However, such studies overlook the im-
plicit influence which the treatment process exercises over the be-
havior under scrutiny (Reid 1992),. Nevertheless, the fact that some
clays targeted in geophagy actually release iron (Johns and Duquette
1991) makes it impossible to rule out this possibility.
20. An additional substance which is widely craved, ice (Knox, Kre-
mer, and Pearce 1995, Cooksey 1995, Smulian et al. 1995, Rainville
1998), has neither of these properties. Hadjadj et al. (1990) suggest
that ice craving may be due to ice’s palliative effects on the gas-
trointestinal consequences of hypoferraemia. However, subjective
accounts emphasize a powerful craving for the substance itself
rather than relief experienced after ingestion (Cooksey 1995). It is
unlikely that an evolved system would motivate ingestion of ma-
terials which have been unavailable for many populations over
most of their history. Crushed ice resembles a mineral powder, and
women report craving ice which is crushed in a particular fashion
fessler Reproductive Immunosuppression and Diet F 33
wash, egg shells, and baby powder, have this effect be-
cause of the presence of calcium; others act through dif-
ferent pathways. For example, amylophagia, the
consumption of indigestible starch, is epidemiologically
associated with hypoferraemia. Although the direction
of causality has been debated (Ephros and Lee 1988), an-
imal models show that starch binds dietary iron, inhib-
iting its uptake (Thomas, Falko, and Zuckerman 1976,
but see also Morais et al. 1996).
Extending an observation of Minturn and Weiher
(1984), Flaxman and Sherman (2000) report that, of the
27 societies listed in the Human Relations Area Files for
which there is information pertaining to gestational nau-
sea and vomiting, the 7 societies in which ethnographers
noted the absence of these symptoms are significantly
more likely to rely on maize as a dietary staple. Corn is
rich in phytic acid, an iron-binding substance that greatly
reduces the bioavailability of dietary iron (cf. Wrobel et
al. 1999). In contrast, polished white rice contains less
than half as much phytic acid as corn (Liu et al. 1993)
and exercises minimal influence on the availability of
iron (Tuntawiroon et al. 1990). Minturn and Weiher
found a positive association between rice consumption
and pregnancy sickness. In Flaxman and Sherman’s as-
sessment, none of the 7 societies reportedly lacking preg-
nancy sickness rely on rice as a primary staple. Among
the Toraja, for whom rice is the primary staple, pregnant
women are reported to experience a powerful aversion
to rice, preferring to eat only corn (Adriani and Kruyt
1996 [1951]). Although it is not clear what the proximate
cues might be, it is possible that pregnant women ex-
perience a strong attraction to corn (but see Pike 1997)
or an aversion to rice (cf. Obeyesekere 1985, Wijewar-
dene, Fonseka, and Goonaratne 1994, O’Brien, Relyea,
and Lidstone 1997:55) or both because the relative ability
of these staples to bind iron somehow influences the
frequency of nausea and vomiting.
Meat is extremely rich in iron, and hence meat eating
poses a double threat to an immunosuppressed woman,
as she risks both ingesting pathogens and supplying them
with the fuel which allows them to multiply within her
gastrointestinal tract. Nevertheless, not all pregnant
women are averse to all meat all of the time. Some of
the variability in the severity and duration of pregnancy
sickness may be due to the existence of heritable differ-
ences with regard to the balance point in the trade-off
between optimal prophylaxis and optimal nutrition.
Likewise, there may sometimes be utility in gambling
by consuming meat. This possibility is enhanced in light
of the centrality of meat in the diets of some hunter-
(Cooksey 1995). Cravings for ice may thus be a misfiring of a mech-
anism aimed at other substances. The consumption of harmful sub-
stances such as paint (Smulian, Motiwala, and Sigman 1995)is
likely the consequence of similar misfirings, and the same probably
applies to many strange olfactory cravings (cf. Cooksey 1995).
21. A reviewer has pointed out that the major iron-binding sub-
stances in the human diet are plant tannins; however, foods such
as tea that are rich in tannins may be avoided because of gustatory
changes (cf. Flaxman and Sherman 2000, but see also Knox, Kremer,
and Pearce 1995).
gatherer groups, making total meat avoidance possibly
very costly in some ancestral populations. How, then,
can the risks of meat eating be reduced? One possibility
is that, although complete decontamination of meat is
unlikely to have been feasible for most of human history,
the same is not true of manipulation of the availability
of iron in the gastrointestinal environment.
Consumption of substances that inhibit iron absorp-
tion significantly reduces the availability of iron from
meat (Thomas, Falko, and Zuckerman 1976). Women
may thus be able to decrease the risks of meat eating by
consuming such substances in conjunction with or soon
after meat. Yet, given that vitamin C has the potential
to enhance iron absorption (and does so independent of
the presence of calcium [Hallberg et al. 1992]), it might
seem that regulation of iron availability in the gastro-
intestinal environment would necessitate a complex
timing mechanism for separating the ingestion of sub-
stances that inhibit the absorption of iron from the in-
gestion of substances that are rich in vitamin C. How-
ever, the need for such a mechanism may be obviated
by the fact that vitamin C has minimal influence on the
absorption of the iron available in meat (heme iron), ex-
ercising its effects primarily on the absorption of the iron
available in plant foods (nonheme iron) (Lynch et al.
1985). Accordingly, a diet rich in both substances that
inhibit iron absorption, such as calcium and indigestible
starch, and foods high in vitamin C may provide addi-
tional protection from meat-borne pathogens while still
meeting the iron requirements of both the mother and
the developing organism.
Problematic Aspects
absence of universality
Flaxman and Sherman analyzed 56 studies involving
79,146 pregnancies and 64,876 women and found a mean
proportion of women experiencing nausea and vomiting
of 62%, with population frequencies ranging from 35 to
84%(2000:119) (see also Emelianova et al. 1999, Weigel
et al. 2000). Ethnographic reviews by Weigel (1985) and
Flaxman and Sherman (2000) have revealed that, though
widely distributed, pregnancy sickness is reportedly ab-
sent in some groups. If it is an evolved adaptation, why
don’t all women experience it?
Pregnancy sickness spans a range of subjective and be-
havorial phenomena, from dramatic bouts of vomiting
to subtle changes in tastes. Studies which discriminate
between nausea and vomiting find that the former is
more pervasive than the latter (Weigel 1985, Tierson,
Olsen, and Hook 1986, Vellacott et al. 1988, Dye, Jones,
and Hill 1998, Zhou, O’Brien, and Relyea 1999, Weigel
et al. 2000). Subtle changes in food preferences are prob-
ably even more prevalent, but because most surveys fo-
cus on nausea and, to an even greater degree, vomiting
they are almost certainly underrepresented in the data
(Profet 1992). Underreporting is probably exacerbated by
the fact that the processes of learning and memory for-
34 F current anthropology Volume
, Number
, February
mation involved in the development of food aversions
are such that many aversions probably shape behavior
in an unconscious fashion (see Bernstein 1999). Lastly,
expectations are an important determinant of the sali-
ence and subjective impact of a variety of aspects of preg-
nancy, and such expectations vary both within popula-
tions and between them (compare Waldenstrom 1999
with Levy 1973; also Mead 1949:21920; O’Brien, Relyea,
and Lidstone 1997). All of these factors can contribute
to variation in the frequency of pregnancy sickness with-
out calling into question the proposition that pregnancy
sickness is an adaptation.
Humans are remarkably flexible, able to exist in a wide
range of environments and subsist on a wide range of
foods. It is likely that this flexibility is ancient, for our
lineage has occupied diverse ecosystems for hundreds of
thousands of years. Even within local settings, seasonal
and year-to-year fluctuations in food availability will
continue to favor dietary flexibility. Nausea, vomiting,
and gestational aversions constrict dietary options and
inflict energetic costs on pregnant women. An optimiz-
ing adaptation can be expected to balance the benefits
of behavioral prophylaxis against a woman’s ability to
bear the attendant costs where that ability is a function
of both the availability of less hazardous alternative
foods and the woman’s overall nutritional status. Con-
sistent with the latter assumption, between-population
comparisons of progesterone levels indicate a positive
relationship with nutritional status (Bentley 1999), sug-
gesting a possible reduction in immunosuppression and
pregnancy sickness under conditions of dietary stress.
More directly, mid-upper-arm circumference and triceps
skinfold thickness—measures of energetic reserves—are
positively associated with the experience of aversions
(Demissie, Muroki, and Kogi-Makau 1998), and there ap-
pears to be a correlation between the severity of nausea
and vomiting and body mass index (reviewed in Huxley
2000, but see also Weigel 1985). Variation in nutritional
status both within and between populations is therefore
likely to contribute to variation in pregnancy sickness,
and it is plausible that variation in the availability of
alternatives to meat will have a similar effect. To date,
investigators have not controlled for such variation, and
hence reported differences in the frequency of pregnancy
sickness shed no light on its utility or lack thereof.
Multiple gestations are associated with increased preg-
nancy sickness (Brandes 1967, Fairweather 1968, Walker
et al. 1985). Multiple gestations can thus contribute to
within-population variation in pregnancy sickness and,
because populations differ substantially in their rates of
twinning (Bortolus et al. 1999), to between-population
variation as well.
Although pregnancy sickness exhibits heritable vari-
ation (reviewed in Flaxman and Sherman 2000), this does
not invalidate the claim that it is an adaptation. Animal
models reveal heritable variation in the propensity to
develop food aversions in response to nausea (reviewed
in Mattes 1991), but there is overwhelming evidence that
this is an adaptation (reviewed in Bernstein 1999). With
regard to both learned aversions and pregnancy sickness,
the goal of avoiding ingestible hazards often conflicts
with the goals of achieving optimal nutrition and opti-
mal foraging efficiency. Because different environments
favor different balance points in this trade-off, gene flow
and/or environmental change may allow for a diversity
of alleles, producing substantial variation in the ten-
dency to prioritize one goal over the other. In sum, ob-
servable variation, a highly overdetermined phenome-
non, does not weaken the argument that pregnancy
sickness is an adaptation.
positive effects of meat consumption on
gestational nausea
Recently, Jednak et al. (1999) found that high-protein
meals, including those containing meat, have a palliative
effect on first-trimester nausea. This finding clearly runs
counter to predictions derived from the meat-avoidance
hypothesis, particularly since the subjects were well-
nourished Western women. However, given the small
sample size (n p 14), this study must be replicated before
definitive conclusions can be reached.
outcome measures
Adaptationist explanations of pregnancy sickness are of-
ten both attacked (i.e. Brown, Kahn, and Hartman 1997)
and defended (i.e., Weigel 1985, Profet 1992, Flaxman and
Sherman 2000, Huxley 2000) on the basis of correlations
between presence or severity of pregnancy sickness and
pregnancy outcome in Western clinical populations.
However, the industrialized nation-state is an evolu-
tionarily novel environment, and it is not clear that the
selective pressures which may have shaped pregnancy
sickness remain unchanged.
In such societies, meat
production, shipment, and storage are closely regulated.
Public education efforts reduce infection resulting from
inadequate hygiene during preparation or insufficient
cooking, and obstetricians routinely warn pregnant
women to handle meat carefully and cook it thoroughly.
Rates of infection proportionate to meat consumption
are therefore doubtless substantially lower than would
have been true in most ancestral populations. It is there-
fore difficult to measure whether pregnancy-induced di-
etary changes offer prophylactic protection against in-
fection by pathogens, and in the absence of extensive
non-Western clinical data inference must be relied upon.
Consistent with the importance of protein intake during
the second and third trimesters, in North America ma-
ternal vegetarianism is associated with low birth weight
and premature delivery (see Sanders 1994). However, in
Karachi, where public health measures are limited in
22. Additionally, there are questions of causality. Although an as-
sociation between lack of pregnancy sickness and increased risk of
first-trimester abortion is marshaled as evidence of pregnancy sick-
ness’s prophylactic effects (Weigel 1985, Flaxman and Sherman
2000), it is unclear whether the problem lies in a deficiency of
pregnancy sickness or in an inadequate shift to anti-inflammatory
activity, causing both termination and an absence of pregnancy
fessler Reproductive Immunosuppression and Diet F 35
comparison, the converse is true, as meat eating during
pregnancy is correlated with lower birth weight and in-
trauterine growth retardation (Fikree et al. 1994), out-
comes associated with many meat-borne pathogens.
cravings for meat
Flaxman and Sherman (2000:127) report that, while meat
is the principal target of aversions, it is not infrequently
a target of cravings as well. If meat is so hazardous that
a special avoidance-motivating mechanism has evolved,
why would it ever be craved? Two mutually compatible
explanations exist. First, as Flaxman and Sherman (2000)
have argued, given the importance of adequate protein
intake, meat may be craved in populations where it is
the principal or even the sole source of protein. The goals
of immunodefense and protein acquisition may be re-
flected in potentially antagonistic aversions and crav-
ings—when alternative protein sources are available,
aversions and cravings do not conflict, but when this is
not the case, one may win out over the other depending
on endogenous cues of need. However, this explanation
alone is insufficient to account for all of the data. The
need for protein is minimal during the first trimester and
maximal during the third trimester, yet Flaxman and
Sherman (2000:127) report that cravings for meat decline
from the first to the third.
Food is a highly marked cultural domain, one in which
symbolic factors independent of utility are likely to play
a role (Rozin 1990). For example, Wijewardene, Fonseka,
and Goonaratne (1994), investigating subjective experi-
ence in pregnant Sri Lankan women, found that 47%
reported craving meat and that cravings across food cat-
egories were more than ten times as common among the
poor than among the rich. Cravings were positively cor-
related with belief in spirit possession and marriage after
a love affair. Sri Lankan spirit possession has been ex-
plained as a means of wish fulfillment, and Sri Lankan
women who marry following a love affair are exhibiting
considerable independence in a world of arranged mar-
riages (Obeyesekere 1981). Accordingly, these two factors
suggest that, at either a conscious or an unconscious
level, these women recognize that they are in a position
to manipulate their social environment. De Silva (1995)
notes that pregnancy is the only time when a Sri Lankan
woman can make demands about her food, and Obeye-
sekere (1985) concludes that Sri Lankan women actively
exploit beliefs about pregnancy cravings so as to control
those who are normally superior to them. He points out
that women crave an array of foods, including those
which are either expensive and rare or taboo to pregnant
women, and he notes that meats fall into both categories
(cf. Wijewardene, Fonseka, and Goonaratne 1994:97)—in-
deed, women may even “crave” nonfood items such as
clothes and jewelry (1985:660).
The pattern described above is not unique to Sri Lanka.
In diverse societies ethnographers record the belief that
the mother and/or child will suffer adverse consequences
if maternal cravings are not indulged.
In both Bengkulu
(my data) and Maryland (Schwab and Axelson 1984),
women explicitly state that pregnancy cravings consti-
tute an opportunity for them to both exercise power over
their husbands and obtain things which are normally
unavailable or forbidden; the same appears to be true
among the Bemba (Richards 1939), Fijians (Thompson
1940), Trukese (Fischer 1998 [1963]); Hokkien Taiwanese
(Barnett 1971), Maasai (Spencer 1988), Javanese (Zeitlin
et al. 1995), Morelos Mexicans (Ingham 1986), Bangla-
deshi (Rizvi 1980), and Saramaka (Price and Price 1991).
Of all food categories, meat is the most universally val-
ued (Stanford 1999). Hence, if any single category is
likely to be the target of manipulative “cravings,” it is
meat (cf. Hanks 2000 [1963], Spencer 1988), and this is
especially likely to be true when meat is scarce and dif-
ficult to obtain (cf. Demissie, Muroki, and Kogi-Makau
1998). Accordingly, the high frequency of cravings for
meat may reveal more about the pervasiveness of gender
inequality (cf. Stanford 1999) than it does about endog-
enous motivations produced by evolved mechanisms.
correlations with progesterone levels
If pregnancy sickness functions to compensate for the
greater vulnerability entailed by reproductive immuno-
suppression, then, given that progesterone plays a key
role in the process leading to reduced maternal immune
activity, we should expect progesterone levels to be as-
sociated with the severity of gestational nausea and vom-
iting. However, empirical results are mixed on this
count. Consistent with the prediction, in addition to be-
ing associated with more severe nausea and vomiting in
comparison with singleton pregnancies, multiple gesta-
tions also involve higher levels of progesterone (Johnson,
Abbas, and Nicolaides 1994) (and, correspondingly, are
associated with greater risk of infection [Mascola, Ewert,
and Eller 1994]). Recent work has focused on the finding
that women experiencing pregnancy sickness exhibit
gastric slow-wave dysrhythmias, and, consistent with
the prophylaxis hypothesis, the same dysrhythmias can
be artificially produced in nonpregnant women through
administration of progesterone (alone or in combination
with estradiol) in doses equal to those occurring during
pregnancy (Walsh et al. 1996). However, in contrast to
these confirmatory results, studies of the relationship
between progesterone levels and severity of gestational
nausea and vomiting report either no correlation (Mas-
son, Anthony, and Chau 1985, O’Connor et al. 1998)or
a negative correlation (Jarnfelt-Samsioe, Bremme, and
Eneroth 1986). These findings constitute a significant
challenge to the prophylaxis hypothesis. Either the hy-
pothesis is wrong or the core hypothesis is correct but
23. For example, Bengkulu (my data), Javanese (Zeitlin et al. 1995),
Maasai (Spencer 1988), Tlingit (Kan 1989), Thai (Hanks 2000 [1963]),
East Toradja (Adriani and Kruyt 1996 [1951]), Ifugao (Barton 1998
[1955]); Morelos, Mexico (Ingham 1986), Hadiya, Ethiopia (Demis-
sie, Muroki, and Kogi-Makau 1998), Bangladesh (Rizvi 1980), Sar-
aguro (Finerman 1985), Tonga (Ostraff et al. 2000), the United States
(Edwards, McSwain, and Haire 1954), Schwab and Axelson (1984).
36 F current anthropology Volume
, Number
, February
the relationship between progesterone and reproductive
immunosuppression is more complex than described
herein. Given the significance of these findings for these
questions, and given that only two of the three studies
confirm one another, further investigation in this area
is vital.
sex during pregnancy
Reproductive immunosuppression presumably increases
vulnerability not only to ingested pathogens but also to
sexually transmitted diseases. Cross-species comparison
among extant primates (Nunn, Gittleman, and Anto-
novics 2000) suggests that, given probable patterns of
human mating, sexually transmitted diseases have likely
been a source of selective pressure throughout our his-
tory. We might therefore expect that, if selection to avoid
pathogens was strong enough to shape ingestion during
pregnancy dramatically, it would also have been strong
enough to favor a reduction in behaviors leading to sex-
ual contact. Consistent with this prediction, pregnancy
is associated with a decline in libido, coital frequency,
and orgasm. However, in stark contrast to the chronology
predicted by the prophylaxis hypothesis, this decline in-
creases successively throughout pregnancy and is most
marked during the third trimester (Kumar, Brant, and
Robson 1981, Hart et al. 1991).
Competing Explanations of Meat Avoidance
As noted earlier, Profet (1992:334) argues that selection
favored gestational meat avoidance because the rapidly
proliferating tissues of the embryo are highly vulnerable
to mutagens and cooked meat contains carcinogenic sub-
stances. North American diet and lifestyle increase the
risk of many cancers. In contrast, investigation of hun-
dreds of deaths among the !Kung, Ache, and Hadza has
not uncovered a single case of cancer despite extensive
meat consumption in each of these societies (N. Blurton
Jones, personal communication). Hunter-gatherers thus
appear not to suffer the consequences that attend meat
eating in North America, presumably in part because
other dietary and lifestyle factors mitigate the carcino-
genic effects of meat. This calls into question the an-
tiquity of selection pressures that Profet associates with
meat eating. Moreover, even in North America, clinical
results suggest that meat consumption during pregnancy
does not increase childhood cancer rates (see Sarasua and
Savitz 1994).
Speth (1991) asserts that selective pressure for a meat-
avoidance mechanism in pregnant women stems from
the fact that, under conditions of restricted caloric in-
take, high levels of protein consumption are dangerous.
Speth notes that (a) most of the calories derived from
24. It is possible that the benefits of reassuring (or confusing) males
as to paternity militate against a reduction in sexual behavior early
in pregnancy, but we must be careful to avoid ad hoc
game are in the form of protein and (b) periodic plant
scarcity drives foragers toward greater meat consump-
tion, increasing the risk of excess protein intake by preg-
nant women. However, the constriction of available re-
sources is significant in proportion to the caloric needs
of the individual—the higher those needs, the more a
scarcity of plant foods is likely to result in dangerous
increases in the caloric contributions of protein. Aver-
sions are concentrated during the first trimester, when
energetic demands are low, and taper off over the second
and third trimesters, as energetic demands climb—the
opposite pattern from that which follows from Speth’s
Other Perspectives on Pregnancy Sickness
Huxley (2000) proposes that first-trimester nausea and
vomiting are an adaptation that reduces caloric intake
in order to stimulate early placental growth. This hy-
pothesis is inconsistent with the concentration of crav-
ings for sweets, fruits, and other high-calorie foods during
the same period. It is likely that enhanced placental
growth is a compensatory response to environmentally
determined scarcity rather than a goal in itself (cf. Ka-
dyrov et al. 1998).
Deutsch (1994) hypothesizes that the evolution of con-
cealed ovulation created selective pressure for a mech-
anism that reduces the frequency of sexual intercourse
during pregnancy in order to protect the developing or-
ganism. This hypothesis is consistent with the risk of
sexually transmitted diseases discussed above. However,
despite the uniquely human nature of gestational vom-
iting, absence of estrus advertisement is found in many
primates (Sillen-Tullberg and Møller 1993). Moreover,
empirical studies indicate that pregnancy sickness does
little to inhibit sexual behavior (Robson, Brant, and Ku-
mar 1981). Deutsch also argues that pregnancy sickness
signals others that the woman requires social support.
However, a woman’s fitness is ill served by a signal that
unnecessarily inflicts costs on herself, her kin, and her
mate, particularly given that pregnancy is accompanied
by cost-free signals such as darkening of the areola and
amenorrhea that are emically recognized under condi-
tions approximating those of ancestral societies (Finer-
man 1985, Shostak 1981, Flaxman and Sherman 2000).
Knox, Kremer, and Pearce (1995), noting that women
who suffer from allergies experience more gestational
aversions than those who do not, argue that such aver-
sions may serve to protect the developing organism from
allergens. This proposal is supported by the observations
that in utero exposure to food allergens does indeed in-
crease the risk of developing an allergy (Frank et al. 1999)
and the propensity to develop allergies has a heritable
component (Rance´etal.1999), suggesting that there
would be benefits to scaling dietary allergen avoidance
in proportion to the mother’s own propensity to develop
allergies. However, if the objective is to shelter the de-
veloping organism from dietary substances that may pro-
voke an immune response, the timing of gestational
fessler Reproductive Immunosuppression and Diet F 37
aversions should parallel the ontogeny of the fetal im-
mune system, yet the exact opposite is true. Further-
more, the patterning of gestational aversions should
closely match the allergenic potential of foods, and this
does not occur with any regularity. While eggs, a com-
mon allergen, are indeed a sometime target of gestational
aversions, peanuts, another common allergen (Rance´et
al. 1999), are not notably avoided. Conversely, beef,
which elicits allergic reactions at only one-fifth the fre-
quency of peanuts (Fiocchi, Restani, and Riva 2000), is
a principal target of aversions.
Profet’s phytotoxin-avoidance explanation of preg-
nancy sickness is neither conclusively supported nor
conclusively disproven by the available evidence. Flax-
man and Sherman (2000) find only intermediate levels
of aversions to vegetables, while my investigation re-
veals that vegetables have minimal salience as a target
of pregnancy taboos. Given both the high concentration
of secondary compounds in spices and the overt nature
of the olfactory and gustatory cues thereof, Profet’s hy-
pothesis predicts strong aversion to spicy foods. Flaxman
and Sherman (2000:127) find only low levels of aversions,
however, and these are matched by equivalent levels of
cravings. Similarly, I found that, although spicy foods
were the only nonmeat category to be tabooed with any
frequency, such taboos were far less common than meat
taboos. Higher rates of aversions and lower rates of crav-
ings are reported for the category “nonalcoholic bever-
ages,” including coffee and tea, but it is difficult to know
what to make of these aversions given that the olfactory
cues that likely play a role in them are not prominent
in nature. In contrast to those mixed results, studies of
gustation strongly support Profet’s hypothesis, as sen-
sitivity to bitterness, the hallmark of phytotoxicity, in-
creases dramatically during pregnancy. Likewise, both
the association between phytotoxin ingestion and pica
across species and the demonstrable toxin-neutralizing
effects of many clays and other inedible substances in-
gested during pregnancy support the assertion that phy-
totoxins may have contributed to the selective pressures
shaping pregnancy sickness.
Following a line of argument that complements my
position and that of Flaxman and Sherman, Haig (1993)
suggests that pregnancy sickness may be a result of ma-
ternal-fetal conflict over the level of maternal invest-
ment. Because immunosuppression puts the mother at