Mueller, U., & Mazur, A. (1998 in press)  Facial Dominance in Homo Sapiens as Honest Signaling of Male Quality. Behavioral Ecology. 

Facial Dominance in Homo Sapiens

as Honest Signaling of Male Quality

Ulrich Mueller
Institute of Medical Sociology
Medical School
University of Marburg
35033 Marburg / Germany
tel: +49-6421-286244
fax: +49-6421-285660
Allan Mazur
Public Affairs Program
The Maxwell School of Citizenship and Public Affairs
Syracuse University
Syracuse NY 13244 / USA
tel: +1-315-445-1970
fax: +1-315-443-5451
accepted for publication in Behavioral Ecology




For a cohort of military officers, graduates of the class of 1950 of the United States Military Academy at West Point, dominant facial appearance was a cosnistent and  important predictor of rank attainment at the academy and - for those who graduated from staff college - for high final rank. For men performing below the average, however, dominant facial appearance was a handicap for promotion. High rank came with high fitness. Thus, facial dominance can be an evolutionarily stable honest signal of dominant behavior, a crucial dimension of the potential for high stastus in a male dominance hierarchy. These findings may apply also to civilian populations.

Key words:   Facial dominance, homo sapiens, honest signaling, rank, reproductive success, status

1 Introduction

The face plays a crucial role in human social cognition (Morris et al. 1996). Among all primates, facial expressions are important signals of internal states: emotions and intentions. Humans also see in the face signals of internal qualities that are more stable over time, like attractiveness as a mate (Barber 1995; Cunningham et al. 1990; Grammer & Thornhill 1994) or dominance (Collins & Zebrowitz 1995; Keating et al. 1981b; Mazur et al. 1984; Mueller & Mazur 1996; Zebrowitz & Montepare 1992; Zebrowitz et al. 1993).

How trustworthy are these signals? It has been shown that honesty of signals, in the sense of allowing receivers to correctly assess the signaler's internal quality, can be evolutionarily stable provided the marginal fitness costs of signaling are higher for individuals with lower quality (Dawkins 1993; Dawkins & Guilford 1991; Enquist 1985; Godfray 1991; Grafen 1990; Grafen & Johnstone 1993; Guilford & Dawkins 1991; Guilford & Dawkins 1995; Hasson 1994; Johnstone & Grafen 1992; Johnstone & Grafen 1993; Maynard Smith 1994). This holds also in noisy environments (Johnstone & Grafen 1992), although a certain degree of dishonesty may then stably exist in the system (Johnstone & Grafen 1993), and it may become advantageous for individuals below a certain quality level not to signal at all (Grafen & Johnstone 1993).

Two kinds of signaling costs may be distinguished (Guilford & Dawkins 1995): investment to produce the signal (for example, the energetic costs of growing a beard), and risk caused by potentially harmful reactions from others (for example, fiercer attacks by competitors). Here we focus on the second kind of cost.

Empirical testing of these theoretical notions is rare in animals (M[emptyset]ller 1987; R[emptyset]skaft & Rohwer 1987; Rohwer 1977) and rarer in humans. Positive relations are observed between high status and conspicuous physical features (height, athletic physique and attractiveness). That may "signal" some underlying fitness enhancing quality. Rarely has a negative impact of such features - as required by theory - been observed, if the individual lacks that quality. In the few studies that demonstrate a negative impact of a conspicuous physical feature with a normally positive impact (Berry & Zebrowitz-McArthur 1988; Zebrowitz et al. 1991), no fitness effects are reported.

Here we show the effect of facial dominance on promotion, for the Class of 1950 of the United States Military Academy at West Point. For men possessing more than minimal cognitive, social and athletic skills, facial dominance predicted promotions, which in turn led to higher fitness. For men lower on these skills, facial dominance was a handicap for promotion. Our interpretation is that marginal costs of signaling vary inversely with the quality we call "potential for high status". We conclude that facial dominance is an honest signal of dominant behavior, which is an important dimension of a man's potential to achieve high status in many social contexts. This is one of the first reported examples of honest signaling in humans, and one of the first empirical studies of honest signaling in any species that includes fitness data.

Data are from public sources and a survey mailed in 1990, when the men, born between 1924 and 1929, had all completed their professional and reproductive careers. More than 80% of the original 670 men were still alive. The class of 1950 was large and successful, producing more generals than other classes in the 1940s and 1950s, but otherwise was typical, making it a suitable object for study. Using an established method (Mazur et al. 1984), individual facial dominance was measured from graduation yearbook portraits on a seven-point scale of dominance-submissiveness by groups of raters. Judges were instructed that a dominant person tells other people what to do, is respected, influential, and often a leader; while submissive or subordinate people are not influential or assertive and are usually directed by others (details in 2.2).

Facial dominance of males of different ethnic background is a signal that respondents from various cultures perceive and interpret in the same way (Collins & Zebrowitz 1995; Keating et al. 1981b). Facial dominance is fairly stable from early adulthood to late middle age (Collins & Zebrowitz 1995; Zebrowitz & Montepare 1992; Zebrowitz et al. 1993), this was confirmed for the sample (Mazur & Mueller 1993).

Facial dominance may signal subjective intentions (Harper 1991; Maynard Smith & Harper 1988) as well as an objective potential for action: For example, strong jaws or broad cheek bones - features that contribute to perceived facial dominance (Cunningham et al. 1990) - may indicate superior physical strength. Possibly they relate to testosterone level (Grammer & Thornhill 1994; Thornhill & Gangestad 1996), suggesting a link to dominant behavior (Mazur & Booth in press). Conversely, people with facial features typical of infants (large eyes, high thin eyebrows, round face, small nose bridges - "babyfacedness") are perceived as warmer, weaker, and more submissive (Berry 1990), and describe themselves as less aggressive (Berry 1991). Progress has been made in identifying other components of facial dominance (Cunningham et al. 1990), but our knowledge is limited about their functional contribution.

We distinguish between dominant behavior, which aims at achieving and maintaining high status and greater control of resources over a conspecific, and aggressive behavior which aims at inflicting physical injury on a conspecific. This distinction is particularly important for humans, because we often assert our status without any intent to cause physical injury. In contexts where the acceptance of high status ultimately relies on the threat of physical force, often the threat is indirect and rarely carried out. There are effective means for securing dominance other than the threat of force. We leave as an important but subsidiary question the extent to which facial dominance may signal aggressive potential.

It is also important to distinguish dominant behavior from its goal: actual dominance or high status. Dominant behavior may be useful for someone aspiring to high status, but usually this goal is achieved only if ego has other social and cognitive skills necessary to obtain from others acceptance of ego´s high status. Without these skills, dominant behavior is perceived as antisocial behavior.

Dominant behavior is an elementary and important social action in all human societies. If respondents reliably rank faces along a dominance-submissiveness dimension, their responses must accord with their experience of men with certain facial features tending to act dominantly. Neurophysiological and neuroimaging evidence shows in primates that certain brain structures specialize in perceiving facial expressions of emotions and intentions, and in regulating emotional and behavioral responses to these expressions (Morris et al. 1996). We suspect a similar biological basis for perception and response to facial dominance.

If perceived facial dominance does indeed accord with potential for high status in male dominance hierarchies, then in the resource based mating system of humans (Buss 1989; Kenrick & Keefe 1992), it signals a fitness relevant quality (Dewsbury 1982; Ellis 1995; Mueller 1993).

The quality "potential for high status" is better observed in the military than other occupations because here, individual power, prestige and standard of living are reliably measurable by formal rank. Also, female-caused reproductive variance may be smaller than elsewhere because military officers, especially in the earlier stages of their careers, live in closed occupational microcosmos, characterized by frequent transfers and housing on military compounds. Personal wealth or outside income have little impact on lifestyle. The wife of a military officer usually has to surrender her civilian career for the institutional nomadism of her husband (Atkinson 1989).

2 Methods

2.1 Data

Data about the professional and family careers of the Class of 1950 were collected from published sources and a mailed survey. West Point's graduation yearbook, The Howitzer 1950, contains cadets' portraits and measures of height, athletic prowess and popularity among peers. The black-and-white portraits are fairly uniform in various aspects known to influence perceived facial dominance: hair was short, and no man wore a beard or a mustache (Muscarella & Cunningham 1996) or glasses (Barber 1995; Mazur et al. 1984). All men wore the same cadet uniform with only the collar and the uppermost parts of the shoulders visible, and without any attributes indicating differential status (Cunningham et al. 1990).

The Academy's annual Register of Graduates and Former Cadets lists class ranking at West Point, and subsequent promotions and command assignments. Following standard research procedures for mailed surveys (Dillman 1979), an autobiographical questionnaire was mailed in 1990 to 539 men, a fairly complete mailing to the survivors of originally 670 men. We received 437 completed questionnaires, an unusually high response rate of 81%. Of 416 classmen who served 20 or more years in the military, 337 (81%) returned questionnaires, including virtually all generals.

Like all cadets, the men of the Class of 1950 entered the military academy as formally undifferentiated plebes, but in their junior year they were ranked as corporals (44%) or privates; in their senior year as cadet officers (25%) or sergeants.

At graduation, all men received the same rank, second lieutenant, and entered one of the army's several branches (e.g., armor, engineers) or the air force, which at that time had no academy of its own.

By 1960, 20% of the class had resigned, for reasons uncorrelated with any variable considered here (Butler 1971). Nearly all of those who remained in the military through the 1950s would stay until 1970 (20 years) or longer in order to retire with benefits.

Promotions and assignments are decided by special boards based on previous assignments and extensive evaluation of personnel records, including recent portraits of candidates. Promotion is nearly automatic to first lieutenant and captain, but early promotion to the rank of major or lieutenant colonel indicates special merit. Between 1960 and 1969, about 75% of the men still on active duty graduated from a staff college, a prerequisite for colonel; between 1962 and 1973, about one half of the staff college graduates graduated from a war college, a prerequisite for promotion to general.

Considering only the 337 respondents who served 20 years or more in the military, the distribution of final ranks is, in ascending order, major (1%), lieutenant colonel (26%), colonel (56%), brigadier general (7%), major general (5%), lieutenant general (3%), and "full" general (2%).

Mandatory retirement years were between 1978 and 1985, depending on rank, though some men left earlier. Median retirement year for lieutenant colonels was 1972, for colonels 1978, for generals 1981. Most men in the sample (98%) married and started to have children (96% of the ever-married) within a few years after graduation from the Academy. Men acquiring illness or disability serious enough to explain differential fecundity would in most cases have resigned from active service. Of the 1057 children born to the 337 men, no child was borne before 1950, and only 28 children were born after 1970. Thus, in our sample, reproduction in most cases was completed well before 1970, many years before retirement, so number of years in military service can be excluded as a causal factor for reproductive success.

2.2 Variables

The modern military, like most other large organizations, bases promotion to high rank on evaluating professional performance and assessing professional potential. Therefore, we measure the quality "potential for high status" by rank attainment and speed of promotion. Indicators of cognitive, social and athletic competence, and of facial dominance at graduation, were used to predict rank attainment, thereby measuring the importance of these factors for success at various stages of the career.

General Order of Merit (GOM). Until 1978, every graduate of West Point was given a number at graduation to indicate his "General Order of Merit" within the class, the lower the better. This aggregate evaluation combines academic grades, peer and instructor ratings of leadership and military aptitude, and physical education grades. At least until midcareer, GOM is known to be related to subsequent promotion (Butler 1976, Mazur et al. 1984).

FRIENDS. There is an approximately 50-word description of each graduating cadet in The Howitzer, typical of student yearbooks. As a crude measure of cadets' sociability, the 41% of cadets whose descriptions made specific reference to their "friends" were scored on the dummy variable FRIENDS.

ATHLETIC. By common standards, nearly all cadets are athletic and well proportioned physically, since these qualities influence admission to the Academy. An ATHLETIC score of 2 (32%) went to men who had been on a intercollegiate sports team, but not in their senior year, in which case a score of 3 (16%) was given, unless they had also won a letter for which they were scored 4 (13%). The rest was scored 1 (36%). This variable measures not just athletic prowess but also competitiveness and ability to cooperate in teams.

FACE. Facial dominance was measured (details in: Keating et al. 1981a, 1981b, Mazur et al. 1984) for all cadets who remained in the military for 20 years or more from their graduation yearbook portraits by 20-40 judges (usually undergraduate classes) who independently rated faces on a seven-point scale of dominance-submissiveness (1 = very submissive, 4 = neutral, 7 = very dominant). Judges were instructed that a dominant person tells other people what to do, is respected, influential, and often a leader; while submissive or subordinate people are not influential or assertive and are usually directed by others. The median score for each slide was taken as the value for FACE. These values ranged from 2 (moderately submissive) to 6 (moderately dominant) with a median value of 4 (neutral) and a mode of 5 (slightly dominant). Smiling, especially broad smiling is known to lower facial dominance: nonsmiling faces scored 0.3 points higher than slight smiles (teeth not visible) and 0.7 points higher than broad smiles (teeth visible), the latter difference being significant on p<.001. Scores of men with broad smile were corrected by adding one point (all relationships involving FACE were checked with corrected and uncorrected scores). Figure 1 displays cadet faces receiving a range of dominance scores.

Figure 1 is unavailable in this electronic version.
Figure 1: Cadets of varying facial dominance who became high-ranked generals, as cadets and at the peak of their careers. Faces are arranged from most to least dominant. These cadets with their ranks and positions in 1983-85 are, from left: Gen. Wallace Hall Nutting (FACE=6), commander-in-chief, U.S. Readiness Command; Gen. Charles Alvin Gabriel (FACE=5), Chief-of-Staff, U.S. Air Force; Gen. John Adams Wickham, Jr. (FACE=4), Chief-of-Staff, U.S. Army; Lt.Gen. Lincoln David Faurer (FACE=3), head of the National Security Agency. Portraits as cadets are from The Howitzer 1950, portraits as officers are from Pentagon archives.

Several potential factors of rank attainment could be excluded from analysis altogether: 1) some features of facial appearance are known to be heritable and, therefore, a relation between face and career might be explainable by persistence of socioeconomic status over generations. Since in our sample measures of parental education including father's military background were unrelated to career (same findings in Butler 1971b, Janowitz 1971) or fecundity, this potentiality need not be considered here. 2) Classmen who entered "combat" branches within the army (infantry, armor), which may attract more dominant personalities, had no different facial dominance scores than other men. Nor were men in these branches more likely to remain on active duty 20 years after West Point. Nor to be war college graduates. Promotion chances were equally distributed over all Army branches. The Air Force offered better career chances, but no variable analyzed here predicted a man's entering the Air Force. 3) Unlike some civilian populations (Frieze et al. 1990; Gillis 1982; Hensley & Cooper 1987), height did not predict any career nor reproduction variable, nor did it correlate with FACE, ATHLETIC or GOM either.

The independent variables GOM, ATHLETIC, FRIENDS, FACE were screened for correlations, using gamma, which is insensitive to marginal distributions. There was no multicollinearity to complicate the interpretation of effects.

2.3 Analysis

We use two approaches (Mueller & Mazur 1996 for technical details) to investigate the dynamics of military careers after graduation from the Academy.

First (3.1.2) we treat the major transitions separately. Single transitions are analyzed with a semiparametric proportional transition rate model with two competing risks: advancement and premature retirement. The latter was uncorrelated with any variable considered here. Because prerequisites of a survival model are only partially fulfilled, we check the transition rate models with the respective logistic regression model.

Second (3.1.3) we investigate all a man's promotions, counting from different base ranks (i.e., lieutenant colonel, colonel, etc.). Log-transformed total numbers of subsequent promotions are analyzed with an OLS regression model.

3 Results

3.1 Professional biography

3.1.1. At the Academy, facial dominance had a clear, monotonic effect on cadet rank in junior and senior year. Percentage of cadet corporals vs. privates rose from the most submissive to the most dominant FACE quartiles: 31%, 42%, 48%, 56% (p<.006), likewise percentage of cadet officers vs. sergeants rose: 9%, 20%, 37%, 46% (p<.002) (details Mazur et al. 1984). GOM, FRIENDS and ATHLETIC were not considered as explanatory variables here, since all three of them may in turn have been affected by cadet rank.

3.1.2. Graduation from staff college and, subsequently, from war college effectively separates officers into three career "channels" (Mazur & Mueller 1996): the first one - no staff college - ending at lieutenant colonel; the second one for staff college graduates without war college ending at colonel; the third one, for graduates of both staff and war college, leading to the top of the military hierarchy. Transitions were analyzed separately for each channel (see Table 1).

Table 1:

Coefficients of proportional hazard rate models (coefficients of logit models in brackets), all covariates standardized.
Risk population:    for staff college graduation: all officers serving 20 years and more;
                            for war college graduates: all staff college graduates;
                            for promotion to any ranks: all officers in the same channel at rank below
Transition to  Staff College Graduation  War College Graduation  Colonel -  Colonel +  Brigadier General +  Major General +  Lieutenant General +  General + 


FACE -.04 (-.28*)  .09 (.14)  .05 (.15)  .08 (§)  .15 (.14)  .26 (.54)  .20 (.51)  .55 (1.20) 
ATHLETIC .11 (.33*)  .16 (.20)  .11 (.43)  .21* (§)  .17 (.21)  .24 (.46)  .33 (.55)  -.49(-.47) 
GOM -.21**(-.51**)  -.23*(-.37**)  -.06(-.02)  -.07(§)  -.15(-.08)  -.06(-.38)  .29 (.02)  -.29(-.56) 
FRIENDS .15* (.39*)  -.01 (-.05)  -.01(-.22)  .21 (§)  .31* (.33)  .34 (.43)  .33 (.44)  .78 (1.03) 
N 321 264 137 136 52 32 17
+ War College Graduates
- Staff College Graduates without War College
§ Virtually all War College Graduates were promoted to Colonel

2-sided tests:
* p < .05
** p < .001


FACE had a negative effect on graduation from staff college. For all those who did graduate, however, FACE is not significant for most single transitions, but the effect (in the transition rate model and in the logistic regression model) is always in the predicted positive direction (sign test: p<.001). For staff college graduates, FRIENDS and ATHLETIC display inconsistent but usually positive effects. GOM, a powerful predictor of graduation from staff and war college, now shows inconsistent effects.

We show survival functions for all six transitions in the channel to general, starting with graduation from war college. We compare men with FACE scores above the median with those at or below. In every rank men with high FACE scores were subject to higher transition rates than men with low scores. Given the strong positive correlation between early promotion and chance of subsequent promotion, the differences between single survival functions understate the real advantage of high FACE scores in this channel. Similarly, in the second channel, men with high FACE scores had a better chance to end up as colonels. In completing staff college, however, such men clearly were at a disadvantage (see Figure 2).

Figure 2: Survival functions for transitions = proportion of men in the respective risk population who have not (yet) experienced the transition, by year
submissive = FACE score at or below median, dominant = FACE score above median.



This latter effect was strong. While facial dominance scores rose from medium performers (staff college graduates without war college) through good performers (colonels with war college) to top performers (generals), those men who did not graduate from staff college, and were not promoted beyond lieutenant colonel, looked as dominant as generals (see Figure 3).

Figure 3: Facial Dominance by Performance Levels:
Figure 3
channel 1: no staff college - lt.colonels
channel 2: staff college, no war college
channel 3: war college: colonels
channel 3: war college - generals

The handicap that dominant looks presented for men apparently lower in professional competence existed not only at this transition, but must have affected their whole career. To illustrate, we use the sum of standardized GOM, FRIENDS and ATHLETIC values (GOM scores multiplied by minus one) as a composite, powerful measure of "career potential". Among men scoring in the lowest 10% on career potential, FACE showed a strong negative correlation with final rank (-.37, p<.03), but among men in the higher 90%, FACE showed a positive correlation with final rank (.14, p<.02).

Also, if an officer did not graduate from staff college, then - with GOM, FRIENDS and ATHLETIC controlled - the more dominant he looked, the later he was promoted to lieutenant colonel, the last rank in reach for him (beta = .39, p < .01). This is the strongest effect of FACE on speed of any transition observed in our data.

3.1.3 Log transformed total number of subsequent promotions was regressed on the independent variables, using four base ranks: lieutenant colonels (staff college graduates), colonel (war college graduates), brigadier general, and major general (see Table 2).

Table 2:

Ordinary least square regressions results (betas and their significance levels) for total number of promotions (log transforms), from base rank lt. colonel: staff college graduates; from higher base ranks: war college graduates only.
Base Rank:  Lieutenant Colonel  Colonel  Brigadier General  Major General 
Indicator Variables: 
FACE .12 
















GOM -.10 

















Model Significance 

n = 














FACE is a significant predictor of the number of promotions from lieutenant colonel, becomes an even stronger predictor of future promotions for colonels and brigadier generals, and remains a sizable predictor of promotions from major general. Figure 4 clearly shows that the more dominant a man's face, the more future promotions he enjoyed. Men with FACE scores of 6 receiving from .3 to .8 more promotions than men with FACE scores of 3, depending on base rank. No men with FACE values below 3 were promoted to general.

Figure 4: Number of subsequent promotions from various base ranks by FACE
Figure 4
Number of subsequent promotions from various base ranks (men with 20+ years of service, for lt. colonel: staff college graduates; for higher base ranks war college graduates only), by FACE. Values for FACE=2, for FACE=6 from brigadier general, and for FACE=3 and FACE=6 from major general are based on very small samples size of 4-9.  Since no officer with a facial dominance score of FACE=2 was promoted beyond colonel, graphs for brigadier and major general start at FACE=3.

FRIENDS is a modest predictor of future promotions from all base ranks. Considering the extreme crudity of this measure of sociability, it is noteworthy that it shows any effect at all. ATHLETIC is a strong predictor of promotions from lieutenant colonel, but only a modest one thereafter. GOM has little predictive power when the promotion pool is limited to war college graduates, although the range of the variable remained almost the same.

3.1.4. Transition rate and logistic regression analysis of single promotions is less powerful than OLS analysis of total promotions, but the effects that do appear are consistent with OLS results. Both point to GOM as an important predictor of graduation from staff and war college but not of later advancement within the pool of war college graduates. Both methods of analysis show FACE to be the strongest negative predictor of staff college graduation, but afterwards FACE was more consistent than ATHLETICS or FRIENDS as a positive predictor of high rank. This effect existed over the whole range of facial dominance scores.

3.2 Reproductive biography

3.2.1. Professional success considerably influenced reproductive success:

Generals had 3.67 children as compared to 3.02 for other officers (p<.001). A closer inspection reveals that the difference in family size is due to a difference in just one parity progression: generals were more likely than non-generals to have a third child ( .94 versus .70 ; p<.0002). This remains if we break down non-generals into war college graduates, staff college graduates, and others (see Table 3).


Table 3:

Total number of children and parity 3 progression by :performance level
total number of children  parity 3 




3.63  .94 

colonels with war college 

2.92  .70 

colonels and lieutenant 
colonels with staff, but 
without war college 

3.10  .70 

lieutenant colonels without staff college 

3.13  .71 
Differences between generals and all other groups in pairwise comparison could be secured on at least p<.03 for total number of children and on at least p<.001 for parity 3 progression. Otherwise, there were no differences between groups in any possible pairwise comparison.

Furthermore, total number of children as well as having a third child, distinguishes winners from losers in all four major transitions: staff college; promotion to colonel in channel 2; war college; and promotion to general (p<.04) (see Table 4).


Table 4:

Parity 3 progression and total number of children by transition outcome:
transition outcome parity 3 

progression: submissive 


parity 3 

progression: dominant 


parity progression 3: 

all face scores 

(total children: 

all face scores) 

staff college  not graduated  .65 .70 .68 (3.04) 
graduated .72 .79 .75 (3.12) 
staff college graduates  lt. colonel  .60 .71 .64 (3.12) 
- no war college  colonel .70 .75 .72(2.98)
war college  not graduated  .67 .75 .70 (3.01) 
graduated .78 .82 .80 (3.22) 
war college graduates  colonel .65 .74 .69 (2.92) 
general .96 .92 .94 (3.65) 
all men  parity 3 progression (total number of children)  .70 





For one-time married men, average waiting time for a third child was longer, and variance of waiting time was greater than for any other child (p<.005). (Maximal family size in the sample was nine). This is not because, in this sample, the third child usually was the last child, for which the waiting time in any case may be longer than for previous children, because also average waiting time for the last child was longer, and variance of waiting time was greater than for any other, if it was a third child (p<.02). The differences between mean waiting times remain, if we include men with more than one marriage.

Generals seemingly were able to invest more in their children than their fellow officers (2.91 vs. 2.38 children with college degree; p<.02 - no difference in birth years), and had more grandchildren at the time of our survey(3.17 vs 2.68, p<.07).

3.2.2. Perhaps an officer with a large family found it easier to be promoted than a man with equal credentials few or no children? However, neither divorce, nor childlessness, nor a small family (one child) - all major deviations from the traditional family ideal - harmed career prospects in any channel at any level, so it seems unlikely that professional success was the effect rather than the cause of above-average fecundity.

Using GOM, FRIENDS, ATHLETIC, and FACE in a multivariate analysis of total number of children, only FACE had a weak positive effect (partial r=.12; p<.04), and also in predicting a third child (partial r=.12; p<.04), which persisted when final rank was controlled. Other things equal, total number of children was higher (3.22 versus 2.93; p<.12), and a third child was more likely (.77 versus .64; p < .05) for men with a FACE score at or above the median. Also, at all four transitions (see Table 4), looking dominant had a positive effect on fecundity.

3.2.3. What about potential confounders? In the sample, most men (98%) married and started to have children (96% of the ever married), within a few years after graduation. Neither divorcees, widowers, nor older men differed in career or reproduction from others. There is no relation between a wife's age, educational background or occupation and her husband's career nor the reproductive performance of the couple. Catholic couples had more children, but religious affiliation was unrelated to any career variable.

4 Discussion

In order to demonstrate that there is selection in favor of honest signaling, we have to show:

1. Facial dominance is a signal of dominant behavior which reflects the internal quality "potential for high status". For individuals with potential for high status, facial dominance should affect rank; otherwise the signal would not have evolved in the first place.

2. The signal "facial dominance" is costly, its marginal cost of signaling greater for individuals with low "potential for high status". As mentioned above, two kinds of signaling costs have to be distinguished: costs from additional investment (a lowering of fitness merely by producing the signal), and costs from additional risk (a lowering of fitness caused by harmful reactions from receivers of the signal). This study only considers costs of the second kind. The costs should be highest for men who signal facial dominance without actually having much potential for high status

3. High status leads to greater fitness.

4.1. Effect of signaling on status attainment.

4.1.1 Facial dominance in the human male is reliably measured by male and female raters all over the world, irrespective of the sender's or the receiver's ethnicity. The basis for this may be common experiences with regard to the elementary and ubiquitious phenomenon of dominant behavior. From recent findings about the perception of, and the response to the facial expression of certain emotions (Morris et al. 1996), it seems possible that there may exist some specific neural basis for the uniform perception of facial dominance among humans.

The development of some features that contribute to perceived facial dominance, for example, strong jaws or broad cheek bones (Cunningham et al. 1990), may be testosterone induced (Grammer & Thornhill 1994; Thornhill & Gangestad 1996). We can regard this findings as indirect evidence of links between dominant looks and dominant behavior (Mazur & Booth in press), independent of ratings.

We conclude, that in the human male, facial dominance is a signal of dominant behavior. The signal need not be perfect; some men act, but do not look dominant, and vice versa.

4.1.2. The modern military in all societies is a dominance hierarchy, where individual behavior is coordinated and controlled by the principle of giving and taking orders. Dominant behavior is important for achieving high rank, but not sufficient. The importance of cognitive, social and athletic skills is highlighted by the effects on career success we observed for GOM (a measure of academic achievement at the Academy), FRIENDS (a measure of sociability) and ATHLETIC (a measure of athletic prowess and also of competitiveness and ability to cooperate within a team).

4.1.3. Signaling dominant behavior paid off, if the cognitive, social and athletic prerequisites were present. Facial dominance while at West Point and on the routes to colonel and general, after graduation from staff college was the most consistent positive factor for career advancement. At the first major filter, the staff college, however, which was not very selective (three quarters of the sample passed), facial dominance was a serious handicap for success.

The stability of facial dominance over a lifetime, the fact that graduation portraits correlate with rank at West Point, as well as 20+ years later, while cadet rank does not correlate with rank in late career (Butler 1976, Janowitz 1971, Mazur et al. 1984), together with the inverse effect for staff college graduation, makes the causal direction behind the statistical relations unambiguous: facial dominance influenced career attainment, not vice versa.

4.1.4. Still, this causal effect of facial dominance on career might be mediated by other mechanisms. Perhaps there is a selection effect (for example Moore & Trout 1978) one could think of a selection effect: Dominant looking men, embodying the stereotype of a military leader, are preferentially selected for West Point, and, as cadets as well as graduates, continued to receive preferential treatment thereafter, irrespective of their actual performance. Such a selection hypothesis has no support in the data:

First, if a selection effect were working, we should expect a positive correlation between FACE and GOM, the overall measure of achievement at the Academy, which was not observed.

Second, smile, in particular broad smile, in many studies has been associated with submissiveness in real interactions (Cashdan 1995; Ellyson & Dovidio 1985). It also decreases the facial dominance score awarded by raters (Mazur et al. 1984), and may have had the same effect on the members of selection boards. We should then expect that successful or ambitious cadets, those who had already attained a cadet officer rank, as well as those who wanted to rise high, were more inclined to present themselves in the graduation yearbook as dominant as possible, and therefore should have avoided posing or selecting portraits with broad smile altogether. Posing as smiling or non-smiling on such occasions seems to be a relatively stable habit: when we investigated portraits of the men in middle age for stability of dominance score over life (Mazur & Mueller 1993), men posing with broad smile on the graduation portraits were posing with broad smile as middle aged men, and vice versa. Smiling while being photographed may reflect an according habit in real interactions. Now, smile was equally distributed over portraits by cadet rank as well as future officer rank, making it unlikely that successful or ambitious cadets purposively tried to look more dominant on official portraits, and, therefore, making it unlikely that facial dominance per se was rewarded.

Third, if facial dominance is valued irrespective of the behavior it is usually associated with, then it should be valued in all candidates and for all promotions alike. By all that we know about socialization in professions and organizations, we should then expect that facial dominance will positively influence promotion chances for all men during early career at least as much as it does later. In this interpretation, there is no room for the severe handicap during early career that dominant looks actually represented for men with low professional potential.

4.1.4..2. Another plausible mechanism behind the causal effect of facial dominance on career success may be male attractiveness (Reis et al. 1982). Perhaps more attractive men, being successful with women, had more successful careers, owing to increased self-confidence.

Facial dominance is an essential part of facial attractiveness in males (Barber 1995; Cunningham et al. 1990; Keating 1985; Krebs & Adinolfi 1975; Sadalla et al. 1987). This also entails some neonate features like large eye height and width, a small nose, expressive features like smile, grooming features (which were controlled in the study sample). Therefore, a dominant looking man may not be rated attractive, and an attractive man need not look particularly dominant. As a consequence, effects of facial dominance and facial attractiveness may be separable by statistical analysis.

We have not had our sample rated for facial attractiveness, but several empirical findings make it unlikely that facial attractiveness and its underlying quality, attractiveness as a mate, are the active components in the observed statistical relations.

First, neither for first nor for second marriages high facial dominance predicted younger wives. Since in all cultures, men prefer their spouses to be the younger, the older they themselves are (Kenrick & Keefe 1992), this can be seen as a hint that facial dominance scores alone did not make men more attractive mates.

Second, for males, smile is the most important expressive feature of facial attractiveness (Cunningham et al. 1990). Since it is known to lower perceived facial dominance (see, it is an ideal criterion for separating facial attractiveness from facial dominance effects. If signaling attractiveness as a mate is the real effective part of the facial dominance signal, then we should expect successful men to smile more frequently than others. But, as mentioned, there was no difference in frequency of smiling on the graduation yearbook portraits between cadet ranks nor between future ranks as officer.

Now, with expressive features irrelevant and variance in grooming features controlled, the facial attractiveness interpretation would lead to the questionable suggestion that it may be exactly the neonate features of their faces that makes dominant looking men rise higher in the military than submissive looking men.

Dominant looking men are attractive mates, precisely because their looks indicate potential for high status. In comparison, other components of facial attractiveness carry less weight, especially if not rating but real behavior is considered. Dominant looks, for example, among male adolescents is a better predictor for early copulation than attractive looks (Mazur et al. 1994).

Thus, we prefer the interpretation that facial dominance has the observed effects on status attainment, precisely because it signals dominant behavior.

4.2. Effect of dishonest signaling on status attainment.

The signal is costly in the sense that signaling a non-existent quality will damage the sender's career more than signaling an existent quality.

We do not need to show that men actually "invest" in the signal, i.e., have their fitness lowered merely by producing the signal. All we have to show is that marginal cost of signaling, as defined above, is inversely related to internal quality. That is, among high quality individuals, signaling should bring advantage as compared to non-signaling competitors (otherwise there would be no selection for signaling in the first place). But signaling should be harmful for low-quality individuals in comparison to non-signaling, in otherwise equal competitors (otherwise there would be no selection against dishonest signaling).

We found the predicted effect in both cases. Men in the talented majority of the sample found it easier to advance, but men in the untalented minority found it more difficult, if they looked dominant. Some of the submissive-looking war college graduates may have made general had they looked more dominant, and some of the dominant looking men who did not graduate from staff college may have passed that filter, had they looked less dominant. The range in measures of cognitive (GOM), social (FRIENDS) and athletic (ATHLETIC) skills was about the same among graduates and non-graduates of both colleges, as well as among generals, although the central tendencies of the respective distributions differed from each other.

4.3. Effect of status on fitness.

4.3.1. The quality "potential for high status" was fitness relevant. Those who made it to general, had about 3.67 children compared to 3.02 children for others and a higher parity 3 progression. There is also evidence that those who passed at other major filters (graduation from staff and war college) had more children and a higher parity 3 progression than those who did not pass (see Table 4). The parity 3 progression from second to third child is known to be the parity progression most sensitive to individual socioeconomic achievement in industrialized societies (Feeney & Lutz 1991; Kravdal 1992). Furthermore, generals were able to invest more in their children, who were more likely to graduate from college, and had more grandchildren than other classmen (3.17 vs. 2.68; proportionally this difference is the same as between number of children). Children of same parity in both groups did not differ in birth years. The children of the classmen were in their peak reproduction years at the time of the survey. Since children with a college degree tend to have their own children later, and since the - larger - families of generals on average are younger than the families of non-generals, the latter difference is likely to increase in the future.

4.3.2. Neither a divorce, nor childlessness nor a one-child harmed career prospects. Therefore, it seems unlikely, that professional success was the effect rather than the cause of above-average fecundity.

4.3.3. Several variables were checked as possible confounders of the observed correlation between the men's professional success and their reproductive performance. In the sample, most men (98%) married within a few years after graduation. Neither divorcees, widowers, nor older men differed in career or reproduction from others. There was no relation between a wife's age, educational background or occupational prestige and her husband's career or the reproductive performance of the couple. Catholic couples had more children, but religious affiliation was unrelated to any career variable. Thus, there is little room for explaining the correlation between men's professional success and their reproductive performance by differences in their marital behavior or in demographic or socioeconomic characteristics of their spouses. We are left with the conclusion that professional success influenced reproductive performance.

Success at the major career filters, however, came years after the respective fecundity transitions, and therefore cannot have directly caused them. In the sample, more than 2/3 of all third children were born before the end of their father's tenth year of service. We do not have good performance measures for that stage of the careers in our data. But in the military as in many other occupational hierarchies, the sorting of men begins early, separating those who will reach the top from those who will not (references in Mazur et al. 1984). Those of our men who after 5 - 10 years in service were set to rise high would have had enough indicators (evaluation records, prestige of assignments, encouragement by mentors) for an assessment of their own career chances. From what is known about the dynamics of third and later births (Bumpass & Westoff 1970) we may speculate, that in many cases the men and their wives decided to have two children as soon as possible, irrespective of the men's military careers. Then, if the career were going fine, they opted for more children. The birth spacing behavior in the sample, with the average waiting time before the third child longer as well as the variance of waiting time greater than before any other child, may be seen as supporting this speculation.

Facial dominance has a moderate effect on reproduction independent of professional success. This is compatible with the theory, for as superiors, peers and subordinates of a man were influenced by the signal, so was his wife.

5 Conclusion

Given the requisite cognitive, social and athletic skills, facial dominance was a crucial factor for career success, which in turn led to higher fitness For those below average on these skills, facial dominance was a handicap for their career.

We conclude that, within the study population, there is a selection in favor of signaling dominance by facial appearance among those who reach high rank, as well as a selection against such signaling in those who fail; in short, a selection for honest signaling.

Because of selective recruitment and socialization, West Point cadets look physically impressive in face, physique and posture, compared to students at other universities, and the same may be said of military officers compared to civilian professionals. Consequently, there is less variation in appearance among military than nonmilitary populations (Collins & Zebrowitz 1995), and the relation between facial dominance and potential for high status - including its effect on fitness - may be even stronger in civilian populations, and so may be the selection in favor of dominant facial appearance as an honest signal of dominant behavior.


The authors appreciate the cooperation of the men of the West Point Class of 1950 and the advice and support of Colonel (Ret.) Morris J. Herbert, United States Military Academy, Association of Graduates. The Office of Institutional Research, United States Military Academy; the US Army Military History Institute and the Center for Air Force History helped us in obtaining portraits shown in figure 1. The research was funded by grant Mu609/5-1 from the Deutsche Forschungsgemeinschaft and by a grant from the Zentrum für Umfragen, Methoden und Analysen (ZUMA) in Mannheim / Germany to the first author.


Atkinson, R. (1989): The Long Gray Line. Boston: Houghton Mifflin.

Barber, N. (1995): "The Evolutionary Psychology of Physical Attractiveness: Sexual Selection and Human Morphology". Ethology and Sociobiology 16, 395-424

Barkow, J.H.; Cosmides, L.; Tooby, J. (1992): The Adapted Mind. Evolutionary Psychology and the Generation of Culture. Oxford University Press.

Berry, D.S. (1991): "Accuracy in Social Perception: Contribution of Facial and Vocal Information.". Journal of Personality and Social Psychology 61, 298-307.

Berry, D.S.; Zebrowitz-McArthur, L.A. (1988): "What's in a Face? Facial Maturity and the Attribuition of Legal Responsibility. Personality and Social Psychology Bulletin 14, 23-33.

Bumpass, L.L.; Westoff, C.F. (1970): The Later Years of Childbearing. Princeton: Princeton University Press.

Buss, D.M. (1989): "Sex Differences in Human Mate Preferences: Evolutionary Hypotheses tested in 37 Cultures". BBehavioral and Brain Sciences 12, 1-49.

Butler, R.P. (1971a): "Official Reasons for Separations from Active Duty of USMA Graduates from the Classes of 1942 through 1966." West Point, NY: Office of Institutional Research, U.S. Military Academy.

Butler, R.P. (1971b): "Survey of Careerists and Non-Careerists From the USMA Classes of 1963-1967" West Point, NY: Office of Institutional Research, U.S. Military Academy.

Butler, R.P. (1976) "Relationships between College Performance and Success as an Army Officer." Journal of Vocational Behavior 9: 385-91.

Cashdan, E. (1995): "Hormones, Sex, and Status in Women." Hormones and Behavior 29, 354-366.

Collins, M.A.; Zebrowitz, L.A. (1995): "The Contributions of Appearance to Occupational Outcomes in Civilian and Military Settings." Journal of Applied Social Psychology 25,  29-163.

Cunningham, M.R.; Barbee, A.P.; Pike, C.L. (1990): "What do Women want? Facialmetric Assessment of Multiple Motives in the Perception of Male Facial Physical Attractiveness" Journal of Personality and Social Psychology 59, 61-72

Dawkins, M.S. (1993): "Are there General Principles of Signal Design?" Phil. Trans. R. Soc. London B 340, 251-255.

Dawkins, M.S.; Guilford, T. (1991): "The Corruption of Honest Signalling". Animal Behavior 41, 865-873.

Dewsbury, D.A. (1982): "Domiance Rank, Copulatory Behavior, and Differential Reproduction." The Quarterly Review of Biology 57, 135-159.

Dillman, D.A. (1978): Mail and Telephone Surveys. The Total Design Method. New York: Wiley.

Ellis, L. (1995): "Dominance and Reproductive Success among Nonhuman Animals: A Cross-Species Comparison". Ethology and Sociobiology 16, 257-333

Ellyson, S.; Dovisio, J. (1985): Power, Dominance, and Nonverbal Behavior. NewYork: Springer.

Enquist, M. (1985): "Communication during Aggressive Interactions with Particular Reference to Variations in Choice of Behaviour". Animal Behaviour 33, 1152-1161.

Feeney, G.; Lutz, W. (1991): Distributional Analysis of Period Fertility. In: Lutz, W. (ed.): Future Trends in Europe and North America. What can we assume today? London, New York: Academic Press.

Frieze, I.H.; Olson, J.E.; Good, D.C. (1990): "Perceived and Actual Discrimination in the Salaries of Male and Female Managers." Journal of Applied Social Psychology 20, 46-67.

Gillis, J. (1982): Too Tall Too Small. Champaign IL: Institute for Personality and Ability Testing.

Godfray, H.C.J. (1991): "Signalling of Need by Offspring to Their Parents". Nature 352, 328-330

Grafen, A. (1990): "Biological Signals as Handicaps". Journal of Theoretical Biology 144, 517-546

Grafen, A.; Johnstone, R.A. (1993): "Why we need ESS Signalling Theory". Phil. Trans. R. Soc. London B 340, 245-250.

Grammer, K.; Thornhill, R. (1994): "Human Facial Attractiveness and Sexual Selection: The Roles of Averageness and Symmetry". Journal of Comparative Psychology 108, 233-242.

Guilford, T.; Dawkins, M.S. (1991): "Receiver Psychology and the Evolution of Animal Signals". Animal Behaviour 42, 1-14.

Guilford, T.; Dawkins, M.S. (1995): "What are Conventional Signals? " Animal Behaviour 49, 1689-1695.

Harper, D.G.C. (1991): "Communication" in: Krebs, J.R.; Davies, N.B. (1991): Behavioural Ecology. Oxford: Blackwell.

Hasson, O. (1994): "Cheating Signals". Journal of Theoretical Biology 167, 223-238.

Hensley, W.E.; Cooper, R. (1987): "Height and Occupational Success: A Review and a Critique". Psychological Reports 60, 843-849.

The Howitzer 1950 West Point, NY: United States Military Academy

Janowitz, M. (1971): The Professional Soldier. New York: Free Press.

Johnstone, R.A.; Grafen, A. (1992): "Error-Prone Signalling". Proc. R. Soc. London B, 229-233.

Johnstone, R.A.; Grafen, A. (1993): Dishonesty and the Handicap Principle". Animal Behaviour 46, 759-764.

Keating, C. (1985): "Gender and the Physiognomy of Dominance and Attractiveness". Social Psychology Quarterly 48, 61-70

Keating, C.; Mazur, A.; Segall, M. (1981a): "A Cross-cultural Exploration of Physiognomic Traits of Dominance and Happiness." Ethology and Sociobiology 2: 41-48.

Keating, C.; Mazur, A.; Segall, M. (1981b): "Culture and the Perception of Social Dominance From Facial Expression" Journal of Personality and Social Psychology 40, 615-626

Kenrick, D.T.; Keefe, R.C. (1992): "Age Preferences in Mates Reflect Sex Differences in Human Reproductive Strategies". Behavioral and Brain Sciences 15, 75-133.

Kravdal, [emptyset]. (1992): "The Emergence of a Positive Relation between Education and Third Birth Rates in Norway with Supportive Evidence from the United States." Population Studies 46, 459-475

Krebs, D.; Adinolfi, A.A. (1975): "Physical Attractiveness, Social Relations, and Personality Style". Journal of Personality and Social Psychology 31, 245-253

Maynard Smith, J. (1994): "Must Reliable Signals always be Costly?" Animal Behaviour 47, 1115-1120.

Maynard Smith, J.; Harper D.G.C. (1988): "The Evolution of Aggression: Can Selection create Variability? " Phil. Trans. R. Soc. London B 319, 557-570.

Mazur, A; Booth, A. (in press): "Testosterone and Dominance in Men". Behavioral and Brain Sciences.

Mazur, A.; Halpern, C.; Udry, J.R. (1994): Dominant looking Teenagers Copulate earlier. Ethology and Sociobiology 15, 87-94.

Mazur, A.; Mazur, J.; Keating, C. (1984)."Military Rank Attainment of a West Point Class: Effects of Cadets' Physical Features." American Journal of Sociology 90: 125-50.

Mazur, A.; Mueller, U. (1993): "Persistence of Facial Dominance Measured from Portraits Taken 20 Years Apart." Manuscript.

Mazur, A.; Mueller, U. (1996): "Channel Modeling: From West Point Cadet to General." Public Administration Review 56, 191-198

M[emptyset]ller A.P. (1987): "Social Control of Deception among Status Signaling House Sparrows Passer Domesticus". Behavioral Ecology and Sociobiology 20, 307-311.

Moore, D.; Trout, B. (1978): "Military Achievement: The Visibility Theory and Promotion." American Political Science Review 72: 452-68.

Morris, J.S.; Frith, C.D.; Perrett, D.I.; Rowland, D.; Young, A.W.; Calder, A.J.; Dolan, R.J. (1996): A Different Neural Response in the Human Amygdala in Fearful and Happy Facial Expressions" Nature 383, 812-815

Mueller, U. (1993). Social Status and Sex. Nature 363, 490.

Mueller, U.; Mazur, A. (1996): "Facial Dominance in West Point Cadets Predicts Military Rank 20+ Years Later." Social Forces 74, 823-850

Muscarella, F.; Cunningham, M.R. (1996): "The Evolutionary Significance and Social Perception of Male Pattern Baldness and Facial Hair". Ethology and Sociobiology 17, 99-118

Register of Graduates and Former Cadets. (volumes 1964, 1980, 1991). West Point, NY: Association of Graduates, U.S. Military Academy.

Reis, H.T.; Wheeler, L.; Spiegel, N.; Kernis, M.H.; Nezlek, J.; Perri, M. (1982): Physical Attractiveness in Social Interaction: II Why does appearance affect social experience? Journal of Personality and Social Psychology 43, 979-996.

R[emptyset]skaft, E.; Rohwer, S. (1987): "An Experimental Study of the Function of the Red Epaulettes and the Black Body Colour of Male Red-winged Blackbirds". Animal Behaviour 35, 1070-1077.

Rohwer, S. (1977): "Status Signaling in Harris Sparrows: Some Experiments in Deception". Behaviour 61, 107-128.

Sadalla, E.K.; Kenrick, D.T.; Vershure, B. (1987): Dominance and Heterosexual Attraction". Journal of Personality and Social Psychology 52, 730-738

Thornhill, R.; Gangestad, S.W. (1996): The Evolution of Human Sexuality. Trends in Ecology and Evolution 2, 98-102

Zebrowitz, L.A.; Kendall-Tackett, K.; Fafel, J. (1991): "The Influence of Children's Facial Maturity on Parental Expectations and Punishments." Journal of Experimental Child Psychology 52, 221-238.

Zebrowitz, L.A.; Montepare, J. (1992): "Impressions of Babyfaceness and Attractiveness Across the Life Span".Developmental Psychology 28, 1143-1152.

Zebrowitz, L.A.; Olson, K.; Hoffman, K. (1993): "Stability of Babyfaceness and Attractiveness Across the Life Span". Journal of Personality and Social Psychology 64, 453-466.