Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://ecology.genebee.msu.ru/3_SOTR/CV_Terekhin_publ/2000_Fecundity_ProcRoySoc.pdf Äàòà èçìåíåíèÿ: Mon Mar 16 10:58:28 2009 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 22:02:01 2012 Êîäèðîâêà:

doi 10.109 8/ rspb.20 0 0.1316

Hu m a n f e r t i l i t y v a r i a t i o n , s i z e -r e l a t e d ob s t e t r i c a l p e r f o r m a n c e a n d t h e e v ol u t i o n o f s e xu a l s t a t u r e d i m or p h i sm
J.- F. Gue gan1*, A . T. Te riok h i n1,2 and F. Thom as1 ¨
re d 'Etude sur le Polymorp hisme des Micro- Organismes, CEPM /UMR CNRS- IRD 9 92 6, Equip e `Evolution des Syste©mes Symbiotiques', IRD, 911 Avenue Agrop olis, B.P. 5 0 45, F- 3 4 03 2 Montp ellier Cedex 1, France 2Dep artm e nt of Bi ology, Moscow Sta te Unive rsity, Moscow 119 8 9 9, Ru ss ia
1Cent

In several animal sp ecies, change in sexual size dimorp hism is a correlated resp onse to selection on fecundity. In humans, di¡erent hyp otheses have been p rop osed to exp lain the variation of sexual dimorp hism in stature, but no consensus has yet emerged. In this p ap er, we evaluate from a theoretical and an emp irical p oint of view the hyp othesis that the extent of sexual dimorp hism in human p op ulations results from the interaction between fertility and size- related obstetric comp lications. We ¢rst develop ed an optimal evolutionary model based on extensive simulations and then we p erformed a comp arative analysis for a total set of 38 countries worldwide. Our op timization modelling shows that size-related mortality factors do indeed have the p otential to a¡ect the extent of sexual stature dimorp hism. Comp arative analysis using generalized linear modelling supp orts the idea that maternal death caused by deliveries and comp lications of p reg nancy (a variable known to be size related) could be a key determinant exp laining variation in sexual stature dimorp hism across p op ulations. We discuss our results in relation to other hyp otheses on the evolution of sexual stature dimorp hism in humans. Keywords : anthrop olog y; fertility ; maternal mortality ; stature ; sexual dimorp hism
1. INTRODUCTION

In all human p op ulation s, men are on average taller than women but the extent of this sexual dimorp hism varies between p op ulations. Although no consensus has emerged to exp lain this variation, several hyp otheses have been suggested. For examp le, well-nourished p op ulations are more sexually dimorp hic than malnourished ones, because male g rowth is more suscep tible to nutritional de¢ciencies during develop ment than female growth ( Hiernaux 19 6 8; Brauer 19 82; Hamilton 19 82). Sexual dimorp hism could also result from natural selection acting di¡erentially on males and females when occup ying di¡erent ecological niches, e.g. di¡erent foraging strategies ( Frayer 19 8 0, 19 81; Brace & Ryan 19 8 0). The `women's work hyp othesis' ( Holden & Mace 19 9 9) suggests that sex-biased p arental investment could be resp onsible for variation in sexual dimorp hism: women would be taller, relative to men, in societies where women contribute more to fo od p roduction because p arents invest relatively more in their daughters, so that the growth of girls would not be comp romised relative to that of boys ( Holden & Mace 199 9). Finally, sexual dimorp hism in stature could result from sexual selection, being greater among p op ulations with p olyg ynous marriage because of intra-male comp etition for females (Trivers 1972; A lexander et al. 1979; but see Gray & Wolfe 19 8 0). Thus, it seems likely that the degree of sexual dimorp hism in human stature may be directly or indirectly in¸uenced by various ecological and p hysiological constraints. In this p ap er we evaluate from a theoretical and an emp irical p oint of view another hyp othesis, that the extent of sexual dimorp hism in human p op ulations results
*

from the interaction b etween fertility (e.g. total number of o¡sp ring born to a woman p assing through child-bearing age) and size- related obstetric comp lications. Human p op ulations are characterized by strong variation in fertility ( Jones 19 9 0). In addition, an imp ressive bo dy of g ynaecological literature has show n that short maternal stature is frequently associated with ser ious obstetric comp lications and often requires an op erative delivery (e.g. Caesarean section or symp hyseotomy ; Camilleri 19 81; Adadevoh et al. 19 8 9; Parsons et al. 19 8 9; Sokal et al. 19 91; Van Roosmalen & Brand 19 92; Tsu 19 92; Kwaw ukume et al. 19 93; Moller & Lindmark 19 97). The correlation between maternal height and obstetrical outcome is so strong that female stature is currently used in most antenatal p rog rammes (e.g. UNICEF, World Health Organization) to screen p regnant women for p otential risk of di¤cult childbirth and cep halop elvic disp rop ortion. Knowing that body height is heritable in humans (e.g. Golden 19 9 4; Carmichael & McGue 19 95; Luo et al. 19 9 8; Arinami et al. 19 9 9; Magarey et al. 19 9 9; Pietilainen et al. 19 9 9), it seems likely that size- related obstetrical p roblems, interacting with variation in fertility, may in¸uence the evolution of female stature and hence sexual dimorp hism. We p redict that in highly fecund p op ulations, selective p ressures for large stature in females are high and sexual dimorp hism is reduced as a consequence. Conversely, in less fecund p op ulations, sexual dimorp hism would be higher because of reduced selective p ressure on females for large size.
2. MATERIAL AND METHODS

(a) Evolutionary op timization modelling
T exp lain the ex istence of sexual stature dimorp hism, and its o dep endence on environmental conditions from the p oint of v iew

Author for co rresp o ndenc e (gue gan@c ep m.mp l.ird.fr).

Proc. R. Soc. Lon d. B (2 0 0 0) 2 67, 252 9^25 35 Rece ived 12 Ju ly 20 0 0 Accep ted 21 Aug ust 20 0 0

252 9

© 2 0 0 0 The Royal Society


2530

¨ J.-F. Guegan and others

Sexual dimorp hism in humans
growth (from birth up to some age of matur ity, T m) and thereafter to rep roduction (after the age of maturity). The body size, W m, reached by the age of matur ity is called the size of matur ity. However, humans continue to grow (up to about 25 years) after the age of maturity (which varies between 12 and 16 years). The e¡ect of p ost- maturity growth emerges in the model if we assume that rep roductive gain dep ends nonlinearly on rep roductive investment, that is, that the function F in equation (3) is nonlinear, namely sigmoid
F (Rt ) ^ Rt /( ¬ Rt ).

of evolutionar y op timality, we have built a state-dep endent model of evolutionarily op timal allocation of energy between growth and rep roduction (Ziolko & Kozlowski 19 83; Kozlowski & T riokhin 19 9 9). We assume that the rate of energ y p roduction e in a unit of time (one year in our model) is p rop ortional (with coe¤cient of p rop ortionality a) to the body mass, wt, at age t to some p ower, b, less than one (e.g. Harvey et al. 19 91), Pt ^ awb . t (1)

This energy is divided between g rowth and rep roduction in p rop ortion ut to 17 ut ; the rate of growth at age t is p rop ortional to the fraction of energy allocated to growth, w
t 1

(8)

^ wt ut Pt ,

(2)

and the rep roductive gain is p rop ortional to some function, F, of the fraction of energy allocated to rep roduction, G
t 1

T comp ute op timal strategies we use the method of stochastic o dynamic p rogramming ( Bellman 1957; Mangel & Clark 19 88). This method consists of two step s. In the ¢rst step, we solve, iterating backwards, the so- called p rincip al equation of dynam ic p rogramming, which in our case can be written in the form Ht ( wt ) ^ maxfH
u
t

^ Gt F (Rt ) ,

(3)

t 1

(w

t 1

) F (Rt )Sexp ( ¡ Q t )g,

(9)

where Rt ^ (1 ¡ ut )Pt . (4)

As a mea sure of evolutionary op timality we use information on the individual's lifetime rep roductive success, calculated as exp ectation of the amount of energy allocated by the individual to rep roduction throughout its entire life from zero to some maximum life sp an, T, which should be su¤ciently large:
T

(we assume HT(wT ) ^ 0 for all wT ). In the second step, we solve, by iterating forwards, the di¡erential equation for body size starting from some initial size, which we set in our comp utations equal to 3.25, i.e. a value close to the weight at birth for a human child.

(b) Comp arative analysis (i) The data set
Data on stature of human males and females were taken from di¡erent p ublished sources and by writing to several ministries of health where national statistics ex isted (our database is available up on request). These data refer to the average human stature (adults aged 18^ 30 years only) for both sexes p er countr y. We collected data on male stature for 72 di¡erent countr ies, and on female stature for 6 6 di¡erent countr ies worldwide. This p ermits the calculation of sexual dimorp hism within p op ulations for 62 di¡erent p op ulations. Since a body of variables may be resp onsible for inter-p op ulation variation in sexual dimorp hism between countries, we considered in our analysis historical (i.e. human ethnic group s), geog rap hical (latitude and longitude), socio - economical ( global net p roduct incomes p er countr y and total, or only vegetable and /or only animal, calorie consump tion p er inhabitant), p arasitological (total number of infectious diseases known in an area), two female life- histor y traits (mean fertility p er female and mean age at menarche) and one obstetrical estimate (maternal-mortality ratio) comp onents. Data for socio - economical p arameters were obtained from the 19 92 world p op ulation data sheet ( Jones 19 9 0): the p er cap ita gross national p roduct ( GNP) in US$ p er year ; calorie consump tion p er average inhabitant p er day, categorized in vegetable-protein consump tion, in animal-p rotein consump tion, or both, taken from food balance sheets from the World Atlas© (19 92). We considered mean latitude and longitude variables, which may a¡ect variation in human stature across countries : the mean latitude and the mean longitude (in degrees) refer to the values measured at the geograp hical centre of each country. Since close geograp hical neighbours may share similar traits, e.g. inhabiting similar environments, simp le cross- country comp arisons are likely to be confounded by distances. The introduction of geograp hical similarities between countries into our generalized linear model alleviates this artefact. Parasites have been shown to p lay a role in the evolution of sexual dimorp hism in many animals (e.g. Hamilton & Zuk 19 82; Andersson 19 9 4),

H0 ^

F( Rt )Lt .
t^1

(5) in our comp utations (the p robability of his age is less than 1076 according to the with typ ical p arameters). in the above equation is de¢ned by , (6)

We set T ^ 110 surviving beyond t Gomp ertz equation The survival, Lt, Lt ^ exp ¡
t¡1 s^ 0

Q

s

where the mortality rate, Q s, at age s is the sum of three terms Q s ^ A Bexp (Cs ) Dexp ( ¡ Ews ). (7)

The ¢rst two term s in this exp ression corresp ond to the equation of Gomp ertz ^ Makeham ( Gomp ertz 1825; Makeham 18 6 0). The ¢rst term do es not dep end on age and may be considered to re¸ect environmentally caused mortality; the second one, which increases with age, may be considered to re¸ect the p rocess of p hysiological ageing. The third term, which we consider to re¸ect adult (acting after age of maturity) size- dep endent mortality (i.e. obstetr ic- related mortality for women, other reasons such a s sexual selection for males), is assumed to decrease exp onentially with increasing body size, ws. Coe¤cient D re¸ects the relative imp ortance of size- dep endent mortality and coe¤cient E re¸ects the e¤ciency of lowering size- related mortality by increasing body size. The p roblem of evolutionary op timization in our ca se consists of searching for the op timal strategy of dividing energy between growth and rep roduction, i.e. in searching for u*(t,w) such that R is max imal for any life history starting with some initial size, w0, and following the strategy u*(t,w). It can be shown, using, for examp le, the Pontryagin maximum p rincip le ( Pontryagin et al. 1962; Ziolko & Kozlowski 1983), that when rep roductive gain and rate of growth dep end linearly on ut the strategy of op timal energ y allocation only has two states: all energy should be allocated to
Proc. R. Soc. Lon d. B (2 0 0 0)


Sexual dimorp hism in humans J.-F. Guegan and others ¨
and so we collected data from both the Centers for Disease Control and the World Health Organisation for a set of 16 categories of human diseases known to a¡ect survival (i.e. typ hoid, hep atitis A, hep atitis B, malaria, schistosomiasis, ¢lariosis, meningococcosis, yellow fever, dengue fever, cholera, tryp anosomiasis, dracunculosis, chagas disease, ly me disease, cutaneous leishmaniosis and visceral leishmaniosis). Based on this information, we calculated the disease load as the total number of disease typ es for each country. We used two di¡erent human life- history traits likely to a¡ect sexual dimorp hism: fertility, which indicates p roductivity as the number of o¡sp r ing born to a woman p assing through child-bearing age, and age at menarche, which rep resents the entry of women into e¡ective rep roductive life. Data on fertility comes from Jones (199 0). T o obtain data on age at menarche, we p erformed a literature search for recent p ublications on this top ic (see Thomas et al. (20 01) for a summary table). The maternal-mortality ratio is de¢ned as the number of maternal deaths (within 4 0 days p ostp artum) caused by deliveries and comp lications of p reg nancy, child-birth and the p uerp erium divided by the number of live births for a given year, and exp ressed p er 10 0 0 0 0 live births. Since this information may di¡er from source to source (e.g. it is not clear whether it includes abortion- related deaths), we used only the statistics of UNICEF in order to have a homogeneous data set (data were available for a subset of 3 8 countries). Furthermore, we categorized the di¡erent p op ulations, according to Cavalli- Sforza (19 97), into eight large divisions of ethnological group s : I, Africans and Nilotics (excep t native p eop le from the Maghreb); II, Europ eans (including p eop le from the Middle East); III, Indians ; IV Mongoloids, Jap anese , and Koreans ; V Amerindians ; VI, New Guineans ^ Pap ous ; , VII, Melanesians ; VIII, Mhongs, Khmers, Thais, Filip inos, Indones ians and related tribes. In addition, a ninth category was assigned to Creoles from Caribbean islands. Ethnological divisions were introduced to control for sim ilarity due to inheritance among related tribes ( Harvey & Pagel 19 91; Martins 199 6). In fact, characters may be inherited a s a result of either behav ioural factors shared by two ethnic group s or by genetic sim ilarity between two given tribes, or both.

2531

dimorp hism, using the p hylogenetic regression p rocedure (one of the most widely used methods of controlling for similarity due to common descent ; Grafen 19 8 9; Martins 19 9 6).

(iv) Generalized linear models
In order to assess the evolution of sexual dimorp hism across countries, we ¢rst analysed the data univariately Then, we used . generalized linear modelling ( McCullagh & Nelder 19 89) to assess simultaneously which exp lanatory variables and / or their interaction terms better exp lained inter-p op ulation di¡erences in stature between the human sexes. For these p urp oses, we used a generalized linear model ( GLM) with a normal error, which rep resents the mo st app rop riate statistical tool for analysing our data. Our max imal model is as follows : sexual dimorp hism ^ a ¸ (log GNP ) + b ¸ (log fert ilit y) + c ¸ (mater nal mortality ratio) + d ¸ (p arasites)+ e ¸ (menarche)+ f ¸ ( longitude) + g ¸ (latitude) + h ¸ (p hylogeny) + (all two- way interactions between the ¢ve ¢rst terms) + ¼. Minimal models were selected with a backward step wise elimination p rocedure. We used the tolerance op tion at the 0.05 level, which avoids constructing highly multi- colinear models in a step wise p rocedure. Here, we used the S- Plus statistical p ackage ( MathSoft, Inc. 19 9 9; Venables & Rip ley 19 9 4). The variances for terms in the model were comp ared using F- tests. When the data suggested there were no linear trends, the exp lanatory variables were transformed and ¢tted again to try to imp rove their contribution to the model. As a result, both GNP and fertility var iables were logarithmically transformed. 3. RESULTS

(a) Evolutionary op timization modelling

(ii) Measuring sexual dimorp hism
For the variation in sexual dimorp hism between p op ulations we used a relative measure equal to the ratio of male size to female size (Alexander et al. 1979; Gray & Wolfe 19 8 0; Gaulin & Boster 19 92). This di¡ers from other measures of sexual dimorp hism used in some p revious studies. In p articular, Harvey & Mace (1982) used residuals from the reduced major ax is since a correlation may exist between the relative ratio of dimorp hism and body size ( Gaulin & Boster 19 85; Ranta et al. 199 4). However, we checked for relationship s between both female and male stature and sexual dimorp hism and they were not signi¢cant ( linear reg ression, r2 ^ 0.0 07, p ^ 0.15 and r2 ^ 0.0 85, p ^ 0.6 8, resp ectively).

(iii) Phylogenetic comp arative analysis
If a human life-history trait is associated with p hylogenetic relatives, this tends to indicate that it is transmitted from ancestor to daughter p op ulations. Thus, two closely related ethnic group s may share similarities transmitted vertically, and therefore treating these ethnic group s as indep endent p oints in statistical analyses may greatly increase the likelihood of typ e I errors. T correct for this p otential bias, we incorp orated p hyloo geny into the statistical analysis of variation in sexual
Proc. R. Soc. Lon d. B (2 0 0 0)

The evolutionary op timization model we described in } 2(a) includes a set of p arameters that must be estimated before running the model. We have accomp lished this estimation using di¡erent app roaches. The values of p arameters B and C in equation (7) were set to typ ical values estimated from demograp hic data : namely, B ^ 0.0 0 0 01 for women and B ^ 0.0 0 0 02 for men and C ^ 0.1 b oth for women and for men (e.g. Gav rilov & Gavrilova 19 91). The p arameter of environmental mortality, A, which varies from 0.0 01 for several Europ ean countries to about 0.013 for several African countries (Thomas et al. 20 0 0) was set to an intermediate value of A ^ 0.0 05. The exp onent b in equation (1) was set to 0.5 on the basis of nonlinear ¢tting of the data p resented in Weinsier et al. (19 92), and falls in the range of values for b typ ically encountered in the literature (e.g. Harvey et al. 19 91). We could not, however, use the estimate for the p arameter a in equation (1) obtained from this nonlinear ¢tting, ¢rst, because we used in our model un its of mass (kg) as units of energy (as can be seen from equation (2)) and not un its of heat (kJ) a s in Weinsier et al. (19 92) and, second, because we did not take into account in equation (1) the energ y allocated to the needs of the organism other than for growth and rep roduction (maintenance, rep air, etc.). We assumed that these other needs constitute an app rox imately constant p rop ortion of all energy p roduced in a un it of time and hence that they simp ly decrease the constant a in equation (1). We estimated this p arameter indirectly, choosing a value of a that leads to p lausible values of women's age and size of maturity after running


¨ 2532 J.-F. Guegan and others

Sexual dimorp hism in humans

80 55 44

(a)

80 70 (b) 55

80 61 50

(c)

w
0

0

13

25

t

110

0

0

15

25

t

110

0

0

14

25

t

110

Fig ure 1. Examp les of evolutionarily op tim al dynam ics of growth fo r di¡erent levels of size-dep endent mortality . ( a) No sizedep endent mortality ( `fem ales') with maturity age t ^ 13 years and m aturity size w ^ 44 kg (w ^ 5 5 kg at t ^ 25 years, age of terminatio n of growth). ( b) Strong size- dep endent mo rtality ( `males') with maturity age t ^ 15 years and maturity size w ^ 55 k g ( w ^ 70 kg at t ^ 25 y ears, age of termination of growth). ( c) Wea k size- dep endent mortality ( `fem ales') with maturity age t ^ 14 years and maturity size w ^ 50 k g ( w ^ 61 kg a t t ^ 2 5 years, age of term inat ion of growth).

the evolutionary op timization model. The estimate of a so obtained was 0.8. Simultaneously, we estimated the p arameters ¬ and in equation (8) as 2 0 0 and 4.5, resp ectively. The results of modelling with this set of p arameters, assuming D ^ E ^ 0, are p resented in ¢gure 1a. We observe growth up to the age of maturity at 13 years, when the body mass reaches 4 4 kg. Growth continues, though at a considerably reduced rate, until 25 years, when the b ody mass becomes equal to 55 kg. Our next step was to show that in the p resence of sizedep endent mortality (D 4 0, E 4 0) op timal size at maturity and adult size increase, and hence that this mortality may exp lain the emergence of sexual dimorp hism during human evolution. Indeed, as we can see in ¢gure 1b, evolutionarily op timal mass at maturity for D ^ 0.01 and E ^ 0.025 is 55 kg (instead of 4 4 kg for D ^ E ^ 0) and adult mass is 70 kg (instead of 55 kg); that is, we obtained values typ ical for men. In a similar manner we can exp lain the reduction of sexual size dimorp hism. For this it is su¤cient to a ssume some size- dep endent mortality in women. Such a situation is illustrated in ¢gure 1c where mass at maturity for D ^ 0.0 01 and E ^ 0.0 25 is 5 0 kg (instead of 4 4 kg for D ^ E ^ 0) and adult mass is 61 kg (instead of 55 kg). These results show that size- dep endent mortality is indeed a factor that can strongly in¸uence evolutionarily op timal size at maturity and adult size. Though we have based the model up on body mass, the above conclusions qualitatively remain valid for body stature because of the strong correlation between human body mass and stature (e.g. Ferembach et al. 19 8 6).
(b) Comp arative analysis

Ta ble 1. Univariate analyses of sexual dimorp hism in human stature against a set of variables that could p ossibly cause intercountry di¡erences
variables fertility ( log ) maternal- mortality ratio gross national p roduct ( log) total calor ie vegetable calor ie anim al calorie p aras ites menarche latitude longitude ethnic group I ethnic group II ethnic group III ethnic group IV ethnic group V ethnic group VI ethnic group VII ethnic group VIII ethnic group IX 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 r .4 .4 .3 .3 .1 .2 .3 .3 .2 .1 .4 .2 .1 .1 .1 .0 .1 .0 .1 27 04 57 17 80 87 15 77 32 29 25 43 21 45 56 69 28 35 56 t 72 .8 72 .6 2.2 2.1 1.1 1.8 71 .9 71 .9 1.8 70 .7 73 .2 1.7 0.8 1.0 1.0 70 .4 70 .8 0.2 1.0 3 5 9 1 5 9 9 9 7 7 4 3 4 1 9 7 9 4 9 5 1 0 4 6 6 2 1 6 9 9 8 3 7 1 8 5 5 5 p (two- tailed) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 .0 .0 .0 .0 .2 .0 .0 .0 .0 .4 .0 .0 .4 .3 .2 .6 .3 .8 .2 0 1 2 4 5 6 5 5 6 4 0 8 0 1 8 3 7 0 7 7 2 8 1 5 5 4 8 9 1 2 9 3 4 1 5 5 7 9

statistically to low sexual dimorp hism across p op ulations. Our model was very robust, accounting for up to 20% of the total deviance.
4. DISCUSSION

In the univariate analysis of sexual dimorp hism across di¡erent human p op ulations, several signi¢cant relationship s ex ist between the extent of stature dimorp hism and the indep endent variables (table 1). However, when p erforming a multivariate analysis, only the interaction term between maternal-mortality ratio and fertility p roved to be a good p redictor of sexual dimorp hism in a step wise generalized linear modelling app roach (table 2 and ¢gure 2). None of the other variables, nor their interactions, entered even in the ¢rst step s of the GLM statistical p rocedure. This imp lies that a high number of maternal deaths caused by deliveries or comp lications of p regnancy, and a high level of birth, both contribute
Proc. R. Soc. Lon d. B (2 0 0 0)

Change in sexual size dimorp hism as a correlated resp onse to selection on fertility is know n in several animal sp ecies (A ndersson 19 9 4; Prenter et al. 19 9 9; Reeve & Fairbairn 19 9 9). Desp ite increasing ev idence that female stature signi¢cantly in¸uences obstetric p erformance, little attention has been devoted to the evolutionary imp lications of this p henomenon. However, our op timization modelling shows that size-related mortality factors indeed have the p otential to a¡ect the extent of sexual stature dimorp hism. First, the action of size-related mortality can exp lain the overall excess of male size over that of females if we assume (e.g. Slatk in 19 8 4; Shine 19 89) that, because of divergence of ecological n iches and


Sexual dimorp hism in humans J.-F. Gue¨ gan and others Ta ble 2 . Summary of GL M for sexual dimorp hism in human stature across 38 di¡erent p op ulations
( Results are obtained after the step wise back ward elimination p rocedure with a which only the interaction term between maternal- mo rta lity rat io and fertility F ^ 8.34 2, p ^ 0.0065). Also given are the p arameter estima te, the standardized p ar degree of freedom in the analysis ( res. d.f.), t- statistics and associated p roba bilities results.) p arameter estimate intercep t maternal mo rtality ¸ fertility 1.079 70.0 01 std. b 0.000 70 . 4 3 4

2533

to lerance op tio n of 0.05. A unique GLM in ( log tra nsfo rmed) was obtained ( r ^ 0.43 4, tial regression co e¤cient ( std. b), the residual . Fo rward stepp ing p rocedure yielded simila r t 455.6 46 72 .888 p ( 4 j t j) 5 0 .0001 0.0065

res. d.f. 38 37

1.10

Libya Greece Jamaica Spain Turkey C hina Poland Bahrain Japan Norway ex-Yugoslavia Ireland Czech R ep. Egypt Singapor N etherlands Thailand e Ecuador France C uba Nicaragua Iran

1.09

Peru

Bolivia

1.08 sexual dimorphism

Cameroon Swaziland Algeria India Papua-New Guinea

1.07

Solomon Is. Tanzania Niger Ivory Coast Kenya Sudan

1.06

C hile

1.05

Fiji

1.04
Namibia

1.03

0

100

200

300

400

500

maternal mortality ´ fecundity Figure 2. Sexual dimorp hism in hum an stature in relation to the intera ction term between maternal- morta lity ra tio and fertility ( log tra nsformed) as p redicted by the GLM. For further exp lanation, see } 2.

rep roductive strategies, the surv ival of males is more directly dep endent on their size. Indeed, our model p redicts that the obser ved sexual size dimorp hism can be exp lained by an excess of the order of 0.0 015 in the annual size- dep endent mortality rate for men above that of women. Second, environmental variation in sizerelated mortality can exp lain changes in sexual dimorp hism. In our modelling, the observed reduction of sexual size dimorp hism can be exp lained by an increase of the order of 0.0 0 01 in the annual size- dep endent mortality for women. Our comp arative analysis using generalized linear modelling supp orts the idea that maternal death caused by deliveries and comp lications of p regnancy (a var iable know n to be size related) could be a key determinant exp laining variation in sexual stature dimorp hism between p op ulations. Sexual dimorp hism in stature was, indeed, negatively correlated with the interaction between maternal-mortality ratio and fertility. This indicates that the variation in dimorp hism dep ends more, on a global
Proc. R. Soc. Lon d. B (2 0 0 0)

scale, on the multip licative e¡ect of the maternal-mortality ratio and fertility than on the sum of the e¡ects of each variable considered sep arately. This is in accordance with our exp ectations, since, to assess the selective p ressure under study, it is necessary to multip ly the risk of dying for a given birth (i.e. maternal-mortality ratio) by the number of birth events during the rep roductive lifetime. Further research would be necessary to assess the imp ortance of this selection during human evolution. However, what seems clear is that women in many countries still exp erience this selective p ressure, for instance in rural areas in A frica where fertility values are among the highest and obstetric interventions are limited or absent. An alternative hyp othesis is that high values of the interaction between maternal-mortality ratio and fertility are a consequence rather than a cause of sexual stature dimorp hism, which would then be determined by other factors. In accordance w ith this idea, Mascie-Taylor & Boldsen (19 8 8) showed, from a large British national samp le, that as husband^ wife height di¡erences increase,


2534

¨ J.-F. Guegan and others

Sexual dimorp hism in humans

so does the p robability of having an abnormal p regnancy outcome. Further investigations would be necessary to a ssess this hyp othesis. For instance, with the help of g ynaecologists, surveys could be conducted in several countries in order to collect data on sp ousal p hysical characteristics and women's rep roductive p erformance. The fact that several variables have not been retained a s exp lanatory variables in the ¢nal model allows us to discuss the relevance of other hyp otheses. For instance, the `women's work hyp othesis' ( Holden & Mace 19 9 9), which stip ulates that sexual dimorp hism is in¸uenced by the amount of fo od received by boys and girls during develop ment (i.e. p arental decisions), is not supp orted here. Under this hyp othesis, rich countries would p robably show lower levels of dimorp hism than p oor countries because p arental discrimination against girls is less frequent in rich countries. Our results indicate the opp osite trend: sexual dimorp hism is higher in rich than in p o or countries (for similar results, see also Eveleth 1975). The `nutrition hyp othesis' ( Hiernaux 19 6 8; Hamilton 19 82), suggesting that well-nourished p op ulations are more sexually dimorp hic than malnourished ones because male g rowth is more suscep tible to nutritional de¢ciencies during develop ment than is female growth, is not supp orted either. Desp ite the higher level of dimorp hism in rich countries than in p oor ones, nutritional variables are not retained in the ¢nal model. However, more investigations are undoubtedly necessary to address this p roblem p recisely. Finally, our study does not supp ort the hyp othesis that sexual- selection p rocesses mediated by p arasites in¸uence sexual dimorp hism. It is frequently argued that comp arative analyses using information from di¡erent sources may be inapp rop riate because data have been collected by di¡erent methods or have come from di¡erent sources. Although this argument is always app licable when no signi¢cant result is detected (i.e. the data are not p recise enough to detect a p otentially signi¢cant result), it is unlikely to be relevant when sig ni¢cant trends are found, since a biological tendency has a p riori no reason to be correlated with backg round noise in the data set ( Mæller 19 97; Lawton 19 9 9). In conclusion, our study supp orts the idea that women are taller, relative to men, in p op ulations where high levels of fertility are likely to counter- select short women because of their higher incidence of obstetrical comp lications. We would like, however, to underline some p o ssible limitations in our study. If the maternal-mortality ratio resp onds very rap idly to recent changes in access to health care, the relationsh ip observed between this p arameter and stature dimorp hism is not op timal. In addition, the sources of env ironmental heterogeneity at the largest scale are very variable. Thus, we must be aware that desp ite our e¡orts to control for such e¡ects, we cannot exclude the p ossibility that other p arameters, at a di¡erent scale, may confound our conclusion.
We are grateful to Sam Brown, Tim Kirkwood, Anders Mæller and Francois Renaud for help ful comments on a p revious draft ° of the manuscrip t. We are also grateful to the four anonymous referees for their constructive comments. F.T. and A.T. are sp onsored by Centre National de la Recherche Scienti¢que, and ¨ J.F.G. by Institut de Recherche p ar le Developp ement.

REFERENCES Adadevoh, S. W., Hobbs, C. & Elk ins, T. E. 19 89 The relation of the true conjugate to maternal height and obstetric p erformance in Ghanaians. Int. J. Gynaecol. Obstet. 2 8, 243^251. Alexander, R. D., Ho ogland, J. L., Howard, R. D., Noonan K. M. & Sherman, P. W. 1979 Sexual dimorp hism and breeding systems in p innip eds, ungulates, p rimates and humans. In Evolutionary biology and human social behavior: an anthrop ological p ersp ective (ed. N. A. Chagnon & W. Irons), pp. 4 02^435. North Scituate, MA : Duxbur y Press. Andersson, M. 199 4 Sexual selection. Princeton Unversity Press. Ar inami, T., Iijima, Y., Y makawa- Kobayashi, K., Ishiguro, H., a Ohtsuk i, T., Y nagi, H., Shimakura, Y., Ishikawa, H. & a Hamag uchi, H. 19 9 9 Supp ortive ev idence for contr ibution of the dop amine D2 recep tor gene to heritability of stature: linkage and association studies. Annls Hum. Genet. 6 3, 147^151. Bellman, R. 1957 Dynamic p rogramming. Princeton University Press. Brace, C. L. & Ryan, A. S. 19 8 0 Sexual dimorp hism and human tooth size di¡erences. J. Hum. Evol. 9, 417^435. Brauer, G. W. 19 82 Size sexual dimorp hism and secular trend: indicators of subclinical malnutrition ? In Sexual dimorp hism in Homo sap iens : a question of size (ed. R. L. Hall), pp. 245^259. New Y : Prager. ork Camilleri, A. P. 19 81 The obstetric signi¢cance of short stature. Eur. J. Obstet. Gynecol. Rep rod. Biol. 12, 347^356. Carmichael, C. M. & McGue, M. 1995 A cro ss- sectional examination of height, weight and body mass index in adult twins. J. Gerontol. A. Biol. Sci. Med. Sci. 5 0, 237^24 4. Cavalli- Sforza, L. L. 19 97 Genes, p eop les and languages. Proc. Natl Acad. Sci. USA 9 4, 7719^7724. Eveleth, P. B. 1975 D i¡erences between ethnic group s in sex dimorp hism of adult height. Annls Hum. Biol. 2, 35^39. Ferembach, D., Susanne, C. & Chambla M.- C. 19 86 L'homme, son e¨volution, sa diversite¨. Paris : CNRS & Doin Editeurs. Frayer, D. W. 19 80 Sexual dimorp hism and cultural evolution in the late Pleistocene and Holocene of Europ e. J. Hum. Evol. 9, 39 9^415. Frayer, D. W. 1981 Body size, weap on use and natural selection in the Europ ean upp er Paleolithic and Mesolithic. Am. Anthrop ol. 8 3, 57 ^73. Gaulin, S. J. & Bo ster, J. S. 19 85 Cross- cultural di¡erences in sexual dimorp hism : is there any variance to be exp lained ? Ethol. Sociobiol. 6, 219^225. Gaulin, S. J. & Boster, J. S. 19 92 Human marriage systems and sexual dimorp hism in stature. Am. J. Phys. Anthrop ol. 8 9, 4 67 ^ 475. Gavrilov, L. A. & Gavrilova, N. S. 19 91 The biology of life sp an: a quantitative app roach. New Y : Harwood Academ ic ork Publishers. Golden, M. H. 199 4 Is comp lete catch- up p ossible for stunted malnourished children ? Eur. J. Clin. Nutr. 4 8, S58^ S70. Gomp ertz, B. 1825 On the nature of the function exp ressive of the law of human mortality and on a new mode of determining life contingencies. Phil. Trans. R. Soc. Lond. A 115, 513 ^ 585. Grafen, A. 19 89 The p hylogenetic regression. Phil. T ans. R. Soc. r Lond. B 326, 19 9^257. Gray, J. P. & Wolfe, L. D. 19 80 Height and sexual dimorp hism of stature among human societies. Am. J. Phys. Anthrop ol. 5 3, 441^456. Ham ilton, M. E. 1982 Sexual dimorp hism in skeletal samp les. In Sexual dimorp hism in Homo sap iens : a question of size (ed. R. L. Hall), pp. 107^163. New Y rk : Praeger. o Ham ilton, W. D. & Zuk, M. 19 82 Heritable true ¢tness and bright birds : a role for p arasites. Scie nce 218, 38 4^387.

Proc. R. Soc. Lon d. B (2 0 0 0)


Sexual dimorp hism in humans J.-F. Guegan and others ¨
Harvey, P. H. & Mace, G. M. 1982 Comp arisons between taxa and adap tive trends : p roblems of methodology In Curre nt . p roblems in sociobiology (ed. King 's College Sociobiology Group ), pp. 343^361. Cambridge University Press. Harvey, P. H. & Pagel, M. D. 19 91 The comp arative method in evolutionary biology. New Y rk : Oxford University Press. o Harvey, P. H., Pagel, M. D. & Rees, J. A. 1991 Mammalian metabolism and life histories. Am. Nat. 137, 556^ 566. Hiernaux, J. 196 8 Variabilite du dimorp hisme sexuel de la ¨ stature en Afrique Subsahar ienne et en Europ e. In Anthrop ologie und humange netik (ed. K. Saller), pp. 42^5 0. Stuttgart, Germany : Gustav Fischer Verlag. Holden, C. & Mace, R. 19 9 9 Sexual dimorp hism in stature and women's work : a p hylogenetic cross- cultural analysis. Am. J. Phys. Anthrop ol. 110, 27^45. Jones, H. 199 0 Pop ulation geograp hy, 2nd edn. London: Paul Chap man Publishing. Kozlowski, J. & T riokhin, A. T. 199 9 Energ y allocation e between growth and rep roduction: Pontr yagin max imum p rincip le solution for the case of age- and sea son- dep endent mortality Evol. Ecol. Res. 1, 423^ 441. . Kwawukume, E. Y., Ghosh, T. S. & Wilson, J. B. 1993 Maternal height as a p redictor of vaginal delivery. Int. J. Gynaecol. Obstet. 41, 27^30. Lawton, J. H. 199 9 Are there general laws in ecolog y ? Oikos 8 4, 177^192. Luo, Z. C., Albertsson-Wikland, K. & Karlberg, J. 19 98 Target height as p redicted by p arental heights in a p op ulation-based study. Pediatr. Res. 4 4, 563^571. McCullagh, P. & Nelder, J. A. 198 9 Generalized linear models, 2nd edn. London: Chap man & Hall. Magarey, A. M., Boulton, T. J. C., Chatterton, B. E., Schultz, C. & Nordin, B. E. C. 19 9 9 Familial and environmental in¸uences on bone growth from 11^17 years. Acta Paediatr. 8 8, 120 4^1210. Makeham, W. M. 18 6 0 On the low of mortality and the construction of annuity tables. J. Inst. Actuaries 8, 3 01^310. Mangel, M. & Clark, C. W. 19 8 8 Dynamical modelling in behavioral ecology. Princeton University Press. Martins, E. P. 19 96 Phylogenies and the comp arative method in animal behavior. Oxford University Press. Mascie-T ylor, C. G. & Boldsen, J. L. 19 8 8 Assortative mating, a di¡erential fertility and abnormal p regnancy outcome. Annls Hum. Biol. 15, 223^228. Mæller, A. P. 19 97 Develop ment stability and ¢tness : a review. Am. Nat. 14 9, 916^932. Moller, B. & Lindmark, G. 19 97 Short stature : an obstetr ic risk factor ? A comp arison of two villages in Tanzania. Acta Obstet. Gynecol. Scand. 76, 39 4^397. Parsons, M. T., Winegar, A., Siefert, L. & Sp ellacy, W. N. 19 89 Pregnancy outcomes in short women. J. Rep rod. Med. 3 4, 357^3 61. Pietilainen, K. H., Kap rio, J., Rissanen, A., Winter, T., Rimp ela, A., Viken, R. J. & Rose, R. J. 19 9 9 Distribution

2535

and heritability of BMI in Finnish adolescents aged 16y and 17y : a study of 48 8 4 twins and 25 0 9 singletons. Int. J. Obes. Relat. Metab. Disord. 23, 107^115. Pontryagin, L. S., Boltyanskii, V G., Gamkrelidze, R. V & . . Mishchenko, E. F. 19 62 Mathematical theory of op timal p rocesses. New Y rk : Wiley. o Prenter, J., Elwood, R. W. & Montgomer y, W. I. 19 9 9 Sexual size dimorp hism and rep roductive investment by female sp iders : a comp arative analysis. Evolution 53, 19 87^ 19 9 4. Ranta, E., Laurila, A. & Elmberg, J. 19 9 4 Reinventing the wheel: analysis of sexual dimorp hism in body size. Oikos 70, 313^321. Reeve, J. P. & Fairbairn, D. J. 199 9 Change in sexual size dimorp hism as a correlated resp onse to selection on fecundity. Heredity 8 3, 697^70 6. Shine, R. 19 89 Ecological causes for the evolution of sexual dimorp hism : a review of the evidence. Q. Rev. Biol. 6 4, 419 ^4 61. Slatk in, M. 198 4 Ecological causes of sexual dimorp hism. Evolution 3 8, 622^ 630. Sokal, D., Sawadogo, L. & Adjibade, A. 19 91 Short stature and cep halop elvic disp rop ortion in Burkina Faso, West Africa. Op erations Research team. Int. J. Gynaecol. Obstet. 35, 347^35 0. Thomas, F., T riokhin, A. T., Renaud, F., De Meeus, T. & e « Guegan, J. F. 20 0 0 Human longev ity at the cost of rep ro¨ ductive success : evidence from global data. J. Evol. Biol. 13, 4 0 9 ^414. Thomas, F., Renaud, F., Bene¢ce, E., De Meeus, T. & « ¨ Guegan, J. F. 20 01 International variability on ages at menarche and menop ause: p atterns and main determinants. Hum. Biol. ( In the p ress.) Tr ivers, R. L. 1972 Parental investment and sexual selection. In Sexual selection and the descent of man 18 71^1971 (ed. B. Campbell), pp . 13 6^179. London: Heinemann. Tsu, V D. 19 92 Maternal height and age: r isk factors for . cep halop elvic disp rop ortion in Zimbabwe. Int. J. Ep idemiol. 21, 9 41^9 4 6. Van Roosmalen, J. & Brand, R. 19 92 Maternal height and the outcome of labor in rural Tanzania. Int. J. Gynaecol. Obstet. 37, 169^177. Venables, W. N. & Rip ley, B. D. 19 9 4 Modern app lied statistics with S- Plus. Berlin: Sp ringer. Weinsier, R. L., Schutz, Y. & Bracco, D. 19 92 Reexamination of the relationship of resting metabolic rate to fat- free mass and to the metabolically active comp onents of fat- free mass in humans. Am. J. Clin. Nutr. 55, 79 0 ^79 4. Ziolko, M. & Kozlowsk i, J. 19 83 Evolution of body size: an op timisation model. Math. Biosci. 6 4, 127^143. As this p ap er exceeds the ma ximum length normally p ermitted, the authors have agreed to contr ibute to p roduction costs.

Proc. R. Soc. Lon d. B (2 0 0 0)