Criminality and the 2D:4D Ratio

Abstract

A decade old theory hypothesizes that brain exposure to androgens promotes involvement in criminal behavior. General support for this hypothesis has been provided by studies of postpubertal circulating levels of testosterone, at least among males. However, the theory also predicts that for both genders, prenatal androgens will be positively correlated with persistent offending, an idea for which no evidence currently exists. The present study used an indirect measure of prenatal androgen exposure—the relative length of the second and fourth fingers of the right hand (r2D:4D)—to test the hypothesis that elevated prenatal androgens promote criminal tendencies later in life for males and females. Questionnaires were administered to 2,059 college students in Malaysia and 1,291 college students in the United States. Respondents reported their r2D:4D relative finger lengths along with involvement in 13 categories of delinquent and criminal acts. Statistically significant correlations between the commission of most types of offenses and r2D:4D ratios were found for males and females even after controlling for age. It is concluded that high exposure to androgens during prenatal development contributes to most forms of offending following the onset of puberty.

Keywords

2D:4D ratio criminality prenatal testosterone neuroandrogenic influences

Only about 20% of criminologists believe that biological factors are contribute significantly to delinquent and criminal behavior (Cooper, Walsh, & Ellis, 2010; Wright et al., 2008). Among the public at large, the percentages appear to be even lower (Campbell & Murcer, 1990; Erskine, 1974; Furnham & Henderson, 1983; Stalans & Lurigio, 1990). Thus, for the vast majority people (criminologists included), the causes of delinquency and crime are confined to the family, societal institutions, and socioeconomic forces. It is therefore not surprising that all the leading theories of delinquency and crime bypass the possibility of biological factors being involved (e.g., Agnew, 1995; Akers, 2009; Arrigo & Young, 1998; Gottfredson & Hirschi, 1990).

One challenge to strict social explanations of delinquency and crime involves the universality of sex differences in offending rates. In other words, no matter how males and females are socially treated, male involvement in serious and persistent crime is always substantially greater (reviewed by Ellis, Beaver, & Wright, 2009). Considering this in conjunction with studies showing that male mammals generally are more aggressive and violent than their female counterparts has led to several proposals that brain exposure to male sex hormones may have a role to play in human criminality (Booth, Granger, Mazur, & Kivlighan, 2006; Ghiglieri, 2000; Volavka, 2002; Udry, 1990).

Evolutionary Neuroandrogenic (ENA) Theory

The most detailed argument for biological factors contributing to criminality has been offered in Ellis’ (2003, 2005) ENA theory. This theory explicitly identifies brain exposure to androgens (so called “male sex hormones”) as increasing the probability of individuals competing with and victimizing others, physically and in terms of confiscating their property. The most influential androgen in this regard is testosterone. It and other androgens appear to fundamentally organize the brain along a masculine-feminine continuum during gestation (Kimura, 1992; Reinisch, 1974; Rogers, 2002). Following birth, most gender differences in brain functioning lie partially dormant until puberty, at which time a resurgence in sex hormones fully activates the brain so that the full extent of gender differences in behavior emerges.

ENA theory asserts that nearly all male brains are exposed to substantially higher levels of androgens than the average female brain. Thus, a central prediction by the theory is that males will exhibit higher levels of competitive/victimizing behavior (of which criminality is purported to be a part), especially following the onset of puberty (Ellis, 2011). While ENA theory can explain the universal tendency for males to be more criminal than females, more direct tests of the theory are hampered by the fact that brain exposure to androgens cannot be precisely measured in a living organism without invasive procedures that damage the nervous system. Thus, most research on the behavioral effects of neurological androgen exposure has been confined to experiments with laboratory animals (Pinna, Costa, & Guidotti, 2005; vom Saal, 1983).

Of course, by definition, only humans commit crime, making this behavior essentially outside the boundaries of experimental animal research. The closest that animal experiments have come to simulating human criminality has involved the study of offensive forms of aggression. These experiments have indicated that brain exposure to androgens promotes aggressive behavior, especially when access to resources or mates is contested (Eising, Eikenaar, Schwabl, & Groothuis, 2001; Rubinow & Schmidt, 1996; Simon & Lu, 2006). While such nonhuman experiments can be considered supportive of ENA theory, their relevance to human criminality is still questionable given that criminality is a uniquely human phenomenon.

Another obstacle to determining whether brain exposure to androgens affects criminality involves timing of the exposure. In mammals, including humans, exposure to androgens occurs over extensive developmental periods. The two most notable periods are called (a) the prenatal (or organizational) stage and (b) the postpubertal (or activational) stage (Arnold, 2009; Romeo, 2003). In line with this two-stage distinction, ENA theory asserts that brain levels of androgens need to be relatively high during both stages to maximize competitive/victimizing tendencies (Ellis, 2005).

Research undertaken to determine whether androgens promote human criminality has been limited to measuring postpubertal testosterone in the blood or saliva. These studies are far from ideal for testing ENA theory because (a) they are nearly always based on a single hormonal assay derived from outside the nervous system and (b) do not assess prenatal androgen levels, only postpubertal levels. Nonetheless, most of these studies have provided support for ENA theory by indicating that postpubertal testosterone levels are positively correlated with rates of offending, at least among males (reviewed by Ellis et al., 2009, pp. 208-210).

Prenatal Androgen Assessment

While it seems likely that prenatal and postpubertal levels of androgens are positively correlated with one another (even within each sex), the evidence is conflicting in this regard (see Voracek, Manning, & Dressler, 2007, for a review and Muller et al., 2011 for additional evidence). Furthermore, twin studies have shown that while circulating postpubertal testosterone levels are highly heritable (Harris, Vernon, & Boomsma, 1998; Kuijper et al., 2007), this may not be the case for prenatal levels. According to one recent twin study, intrasex prenatal testosterone levels were almost entirely the result of intrauterine environmental factors rather than genetics (Caramaschi, Booij, Petitcherc, Boivin, & Tremblay, 2012).

Findings of this nature suggest that just because criminality appears to be positively associated with postpubertal androgen levels does not mean that the same is true for prenatal androgens. Therefore, the ENA hypothesis that prenatal androgens will be positively associated with involvement in delinquency and crime needs to be empirically tested.

Beginning in the late 1990s, researchers began to discover evidence that prenatal androgen levels could be inferred from the measurement of a rather simple physiological trait known as the 2D:4D finger ratio (Fink, Thanzami, Seydel, & Manning, 2006; Manning, Scutt, Wilson, & Lewis-Jones, 1998). To gauge this trait, one simply assesses the relative length of the second (index) and fourth (ring) fingers. The longer the fourth digit is compared with the second—that is, the lower the ratio—the greater the exposure to prenatal androgens (Hönekopp, Manning, & Muller, 2006; Lutchmaya, Baron-Cohen, Raggatt, Knickmeyer, & Manning, 2004).

As to how such a finger length ratio would be affected by perinatal androgens, it is well known that testosterone promotes bone growth (Ebeling, 2010; McIntyre, Ellison, Lieberman, Demerath, & Towne, 2005; Sims, Brennan, Spaliviero, Handelsman, & Seibel, 2006). One of the critical periods when androgens normally rise prenatally appears to coincide with a growth spurt in the fourth digit (Brown, Hines, Fane, & Breedlove, 2002; Lutchmaya et al., 2004; Malas, Dogan, Hilal Evcil, & Desdicioglu, 2006). Thus, the relative length of the index and ring finger seems to provide at least a rough indication of androgen levels during fetal development.

There are two noteworthy caveats to the points just made. First, numerous studies have indicated that the 2D:4D ratio on the right hand is superior to that on the left hand for estimating prenatal androgen exposure (Garn, Burdi, Babler, & Stinson, 1975; Manning et al., 1998; Williams et al., 2000). Second, the hormones and their prenatal timing regarding digit growth are not identical to what affects masculinization/defeminization of the brain (Berenbaum, Duck, & Bryk, 2000; Sharpe, 1998). Therefore, the 2D:4D ratio cannot be considered more than an approximate peripheral biomarker for the extent to which the brain was affected by androgen exposure (e.g., Hönekopp, Manning, et al., 2006; Hönekopp, Voracek, & Manning, 2006; Lutchmaya et al., 2004). This implies that only with large samples can one expect to obtain consistent findings regarding behavioral traits associated with 2D:4D (Breedlove, 2010).

Evidence That 2D:4D and Offending Are Associated

So far, evidence for an association between the 2D:4D ratio and offending is quite limited. In terms of indirect evidence, one study reported that the 2D:4D ratio was lower (i.e., indicating higher prenatal androgens) among persons who scored high in psychopathy (Blanchard & Lyons, 2010). The ratio was also low among males who were unusually high in aggression and risk taking (Hönekopp, 2011). However, another study failed to confirm this pattern for risk taking (Anderson, 2012).

Regarding law violating behavior more specifically, three studies were located. One indicated that among a sample of 77 males, those with the lowest 2D:4D finger ratios had accumulated more moving traffic violations than males as a whole (Schwerdtfeger, Heims, & Heer, 2010). Another study compared the finger ratios of 44 males with at least one criminal conviction and 46 male with no convictions. It revealed that the convicted males had significantly lower 2D:4D ratios (Hanoch, Gummerum, & Rolison, 2012). However, another study compared the 2D:4D ratio of 44 male convicts with 66 males with no conviction record. It found no significant differences between these two groups (Anderson, 2012).

Purpose of the Present Study

The present study was undertaken with two large ethnically diverse samples to determine whether the 2D:4D ratio on the right hand (henceforth referred to as r2D:4D) is inversely associated with self-reported involvement in criminal behavior. The theoretical expectation is that to the extent that the r2D:4D ratio is indicative of prenatal brain exposure to androgens, negative correlations should exist between this ratio and involvement in delinquent and criminal behavior.

Method

Data for this study were obtained from respondents living in two distinct countries: Malaysia and the United States. The Malaysian sample consisted of 2,059 undergraduate college students enrolled at the University of Malaya in Kuala Lumpur. U.S. respondents were undergraduates attending the following seven universities: Boise State University (76 respondents), California State University at Fullerton (251 respondents), Evangel University (269 respondents), Minot State University (149 respondents), Pennsylvania State University (110 respondents), University of Missouri (258 respondents), and University of Texas at San Antonio (178 respondents), for a total of 1,291.

Basic demographic characteristics of the two samples are shown in Table 1. As one can see, the gender and age distributions were very similar. It is worth noting that the greater proportion of females in Malaysia and the United States most likely reflects the fact that substantially more females than males are attending college in both countries in recent years (DiPrete & Buchmann, 2006; Firebaugh & Dorius, 2010).

Table 1.

Basic Demographics.

Demographics	Malaysian sample	United States sample
Gender
Males	653 (31.7%)	472 (36.6%)
Females	1,406 (68.3%)	819 (63.4%)
Total	2,059	1,291
Age
M (SD)	20.87 (2.36)	20.17 (3.58)
Range	18-42	17-57
Total	2,029	1,291
Marital status
Single	1,971 (95.7%)	1,205 (93.3%)
Married	37 (1.8%)	67 (5.2%)
Divorced/separated	1 (0.0%)	13 (1.0%)
No response	50 (2.4%)	6 (0.5%)
Total	2,059	1,291
Ethnicity
White/European Ancestry	0	859 (67.0%)
Black/African Ancestry	0	81 (6.3%)
Hispanic/Latin/Native American	0	199 (15.5%)
Malay/Bumiputera/Indonesian	1,470 (71.4%)	4 (0.3%)
East Asian (Chinese, “Asian” in United States)	477 (23.3%)	53 (4.1%)
Other Asian (primarily Indian)	88 (4.3%)	43 (3.4%)
Other (Mixed, Arabic, Persian, Euro-Asian)	4 (0.2%)	44 (3.4%)
No response	20 (0.9%)	8 (0.6%)
Total	2,059 (V)	1,291 (V)

Regarding ethnicity, the Malaysian and U.S. student samples were striking different. In particular, whereas 67% of the U.S. students were Whites (i.e., of European ancestry), none of the Malaysian students were. Instead, the majority of the Malaysian respondents described their ethnicity as native Malays (including Bumiputeras, roughly translated to mean “sons of the soil”). Nearly all of the roughly 30% who did not classify themselves as native Malays described themselves as being of either Chinese or Indian ancestry.

The Questionnaire

The data were derived from a questionnaire developed and refined in English. Then it was translated into the native Malaysian language, Bahasa Malaysia. To ensure that the translation carried the same meaning as the original English version, the latter questionnaire was back-translated into English until all discrepancies were eliminated. Both questionnaires were four pages in length and covered a wide variety of topics, only a few of which are the focus of the present study.

Androgenic Exposure Measures

As documented in the introduction, the length of the pointing finger (2D) relative to the ring finger (4D) has been shown to be a proxy measure of prenatal exposure to androgens. Specifically, the greater the 2D:4D ratio, the lower the exposure. Furthermore, as noted earlier, the ratio appears to be more reliably exhibited on the right hand than on the left (Falter, Plaisted, & Davis, 2008). Consequently, in the present study, only the ratio for the right hand was measured.

The precise wording of the question used to measure r2D:4D was as follows: Hold up the back of your right hand so that you can see all five fingers. With your thumb as the first finger, compare the lengths of your second (pointing) finger with your fourth (ring) finger. Which is longer? (check one of the five responses below)

Pointing (2nd) finger considerably longer

Pointing (2nd) finger slightly longer

They are almost exactly the same length

Ring (4th) finger slightly longer

Ring (4th) finger considerably longer

The above five responses were coded in reverse order from 5 to 1, respectively, thus rendering an interval measure of the r2D:4D ratio with lower ratios indicating the highest exposure to androgens.

As an indicator of the validity of the r2D:4D measure used in this study, we ran t-tests of the sex differences for each of the racial/ethnic groups that were substantially represented in both countries. Table 2 shows that as expected, in all five groups, the r2D:4D ratios were significantly higher in females than in males. This is consistent with findings from other researchers (Brabin, Roberts, Farzaneh, Fairbrother, & Kitchener, 2008; Manning, Stewart, Bundred, & Trivers, 2004).

Table 2.

Mean Sex Differences t-Test for the r2D:4D Ratios for the Five Main Racial/Ethnic Groups.

Ethnic group	Male mean	Female mean	T-ratio	Cohen’s d
White/European Ancestry	2.238	2.608	6.18***	.389
Black/African Ancestry	2.088	2.603	2.83**	.571
Hispanic/Latin/Native American	2.280	2.655	3.59***	.382
Malay/Bumiputera/Indonesian	2.620	2.842	3.58***	.202
East Asian	2.401	2.860	4.69***	.434
Total sample	2.406	2.766	10.06***	.346

p < .05. **p < .01. **p < .001, two-tailed test.

To help assess the reliability of our self-report r2D:4D measure, we collected new data in which caliper-based measurements were compared with the self-report measurement herein reported. The main comparison involved 215 U.S. college students whose right hand was scanned palm down on a desktop scanner/printer. The other comparison involved 35 U.S. students who provided an outline sketch around their right hand (using a pen with their left hand) on a blank sheet of paper. The results were as follows: For, the scanned measure, the correlation with the five-category self-estimate was r = .30, while for the hand outline measure, the correlation with the five-category self-estimate was r = .44.

We were surprised that the correlation between the hand outline and the self-report measures was higher than the correlation involving the scanned measure, but are inclined to attribute it to the small sample size obtained with the latter comparison. Nevertheless, the main point is that both correlations indicate that at least for U.S. respondents, the five-category self-assessed r2D:4D finger length measure used in the present study was at best only modestly reliable. Assuming that the error from this self-assessed measure is random, one can consider findings from this study as providing low estimates of the correlations that would have been obtained with more precise r2D:4D measurement (Allaway, Bloski, Pierson, & Lujan, 2009; Kemper & Schwerdtfeger, 2009).

A caliper-based measure was not used in the present study for three reasons: First, the questionnaire was designed by a consortium of researchers in two countries with the understanding that the final version would not exceed four pages, thus restricting the number of variables that could be measured. Second, only a minority of the researchers cooperating in data collection were interested in the 2D:4D variable, thereby minimizing the space that could be allocated to it. Third, had the respondents been asked to provide a printed scan of their hands, it would have been difficult to assure them of anonymity, rather important because they were asked to report involvement in illegal behavior.

To assess delinquent and criminal behavior by the respondents, the research consortium responsible for the data agreed on asking about behavior that would be considered (a) illegal, (b) reasonably common, and (c) easily understood in both countries. The results were the following 13 acts:

Serious assault or beating (needing medical treatment)

Minor assault or beatings (not needing medical treatment)

Sexual assault (including attempted sexual assault, molesting, rape)

Domestic or courtship violence

Reckless driving (e.g., road bullying)

Serious theft or robbery (including motor vehicle theft)

Minor theft or robbery (including shoplifting, purse snatching, pick pocketing)

Serious damage to property (major vandalism and arson)

Minor property damage (minor vandalism such as breaking windows in houses or cars)

Bribery, fraud, or other finance-related offenses

Distribution of illegal drugs (excluding alcohol)

Use of illegal drugs (excluding alcohol)

Illegal gambling

The specific number of each offense that respondents stated they had committed was recorded unless it exceeded 99, in which case 99 offenses were recorded.

As a closing item on the questionnaire, respondents were asked the following question: How carefully did you read and try to honestly answer the questions in this questionnaire? Responses to this question as well as each respondent’s gender and age were used in the analysis.

Data Analysis

For assessing the basic relationships between r2D:4D and criminal involvement, we used Spearman’s rho. The statistical technique adopted for the multivariate analyses was negative binomial regression. This type of regression is appropriate for models that use count data for the dependent variable that have few nonzero cases and that are overdispersed (Hilbe, 2011; Kremelberg, 2011). An overdispersed measure has a standard deviation that is larger than its mean. This feature was characteristics of all 13 crime measures.

Results

Results from correlating r2D:4D with each of the 13 forms of criminal behavior are presented in Table 3. This table reveals that as theoretically predicted, 35 out of the 36 coefficients were negative, although only 17 were statistically significant. When data for the two countries were considered separately, seven of the eight statistically significant correlations involved the Malaysian sample, while only one was significant in the U.S. sample, likely due to the Malaysian sample being roughly 40% larger. Combining data for both countries together resulted in all but four of the correlations being statistically significant. In descending order, the following types of offenses were most strongly associated with reduced r2D:4D ratios: reckless driving, illegal drug use, gambling, major vandalism, finance-related offenses, minor vandalism, minor assault, illegal drug distribution, and serious assault.

Table 3.

Spearman Rank Correlations Between r2D:4D and Self-reported Offending for 13 Offense Types for Malaysian Students and United States Students Separately and Together.

Offense types	Spearman rho
Offense types	Malaysia	United States	Combined
Serious assault	−.039*	−.012	−.030*
Minor assault	−.071**	−.009	−.052**
Sexual assault	−.023	−.052	−.037
Domestic assault	−.036	.039	−.005
Serious theft	−.037	−.026	−.030
Minor theft	−.047*	.022	−.028
Major vandalism	−.029	−.009	−.061***
Minor vandalism	−.041*	−.036	−.056***
Reckless driving	−.054**	−.034	−.075***
Bribery and fraud (financial crime)	−.039*	−.073*	−.056***
Illegal commerce (mainly drugs)	−.019	−.006	−.031*
Illegal drug use	−.034	−.030	−.068***
Illegal gambling	−.056**	−.059	−.066***

p < .05. **p < .01. ***p < .001, one-tailed test.

To explore how criminal behavior is related to r2D:4D when other key variables were entered into the equation, we performed a series of negative binomial regression models according to all 13 categories of crime. Table 4 displays results for serious and minor assault. One can see that low r2D:4D ratios were significantly related to involvement in serious assaults for both genders, but that minor assaults were not.

Table 4.

Negative Binomial Regression Model Using r2D:4D Digit Ratio, Sex, Age, and Honesty as Predictors of Two Categories of Assault.

Predictors	Serious assault			Minor assault
Predictors	Total sample	Males	Females	Total sample	Males	Females
r2D:4D ratio	−.35** (.13)	−.35** (.16)	−.38* (.23)	.05 (.05)	.08 (.05)	.04 (.07)
Male	1.55*** (.26)	—	—	1.54*** (.10)	—	—
Age	.04 (.03)	.06 (.04)	−.02 (.08)	.05** (.01)	.04* (.02)	.05** (.02)
Honesty	.44**(.13)	.37** (.14)	.68* (.32)	.13*** (.03)	.17*** (.05)	−.001 (.05)
Constant	−8.50*** (1.50)	−6.80*** (2.51)	−9.63** (3.47)	−4.79*** (.45)	−4.11*** (.61)	−3.65*** (.64)
Model χ²	73.78***	17.49***	11.19**	300.71***	34.14***	6.13
n	3,159	1,055	2,104	3,159	1,055	2,104

Note. Standard errors are presented in parentheses.

p < .05. **p < .01. ***p < .001, one-tailed test.

Table 5 has to do with sexual assault and domestic assault. It reveals substantial effects of r2D:4D on sexual assault for males and females, but no effects on domestic assault among either gender, possibly due to low involvement in domestic violence among college samples. As with serious assaults, low finger length ratios were significantly associated with sexual assaults self-reported by males and females.

Table 5.

Negative Binomial Regression Model Using r2D:4D Digit Ratio, Sex, Age, and Honesty as Predictors of Sexual and Domestic Assault.

Predictors	Sexual assault			Domestic assault
Predictors	Total sample	Males	Females	Total sample	Males	Females
r2D:4D ratio	−.79** (.30)	−.65* (.31)	−1.99* (1.21)	.01 (.11)	−.12 (.18)	.07 (.15)
Male	2.72*** (.75)	—	—	.64** (.23)	—	—
Age	−.48* (.19)	−.65** (.18)	.05 (.13)	.13*** (.02)	.16*** (.02)	.11*** (.03)
Honesty	−.45*** (.08)	−.45*** (.09)	−.50* (.25)	.06 (.08)	−.12** (.09)	.33* (.15)
Constant	8.17** (3.97)	−13.83** (4.41)	.05 (3.59)	−7.32*** (.84)	−5.30 (.94)	−9.49*** (1.61)
Model χ²	69.14***	35.53***	7.65*	69.38***	47.52***	20.18***
n	2,896	985	1,911	3,158	1,054	2,104

Note. Standard errors are presented in parentheses.

p < .05. **p < .01. ***p < .001, one-tailed test.

Results concerning three categories of property crimes are shown in Table 6. There, one can see that serious thefts as well as financial crimes (such as bribery and fraud) were inversely related to r2D:4D ratios. However, finger ratios were only significantly associated with male involvement in minor thefts, not female involvement.

Table 6.

Negative Binomial Regression Model Using r2D:4D Digit Ratio, Sex, Age, and Honesty as Predictors of Three Categories of Property Crime.

Predictors	Serious theft			Minor theft			Financial crime
Predictors	Total	Males	Females	Total	Males	Females	Total	Males	Females
r2D:4D ratio	−.24* (.14)	−.32 (.18)	−.13 (.22)	−.05 (.04)	−.18** (.06)	.05 (.05)	−.68*** (.12)	−.25* (.13)	−1.66*** (.24)
Male	1.31*** (.27)	—	—	.84*** (.08)	—	—	1.04*** (.19)	—	—
Age	−.12 (.07)	−.06 (.08)	−.25 (.15)	.003 (.01)	.07*** (.02)	−.16*** (.03)	−.001 (.03)	.04 (.04)	−.07 (.06)
Honesty	.15 (1.01)	.03 (.10)	.65* (.31)	.33*** (.04)	.23*** (.05)	.45*** (.06)	.04 (.06)	.00 (.08)	.16 (.13)
Constant	−2.99 (1.81)	−1.48 (1.92)	−5.44 (4.45)	−4.50*** (.43)	−3.79*** (.55)	−2.72*** (.84)	−2.50** (.94)	−2.75** (1.10)	−.40 (1.78)
Model χ²	38.46***	4.58	12.43**	236.72***	61.57***	125.54***	85.96***	4.78	71.82***
n	3,159	1,055	2,104	3,159	1,055	2,104	3,159	1,055	2,104

Note. Standard errors are presented in parentheses.

p < .05. **p < .01. ***p < .001, one-tailed test.

Moving to offenses involving intentionally causing property damage (vandalism), Table 7 shows that serious damage was significantly related to r2D:4D only for females, while minor damage was so related only in the case of males. The direction of both of these significant associations were once again inverse.

Table 7.

Negative Binomial Regression Model Using r2D:4D Digit Ratio, Sex, Age, and Honesty as Predictors of Two Categories of Vandalism.

Predictors	Serious property damage			Minor property damage
Predictors	Total sample	Males	Females	Total sample	Males	Females
r2D:4D ratio	−.21* (.11)	−.08 (.12)	−.68** (.26)	−.01 (.04)	−.09* (.05)	.09 (.06)
Sex	1.91*** (.25)	—	—	1.40*** (.08)	—	—
Age	−.02 (.04)	.02 (.04)	−.42* (.18)	.004 (−.01)	.04** (.02)	−.08** (.03)
Honesty	−.14*(.05)	−.14* (.06)	−.15 (.12)	.15*** (.03)	.20*** (.04)	.07* (.04)
Constant	−2.49* (1.07)	−1.73* (1.04)	−.6.59* (3.86)	−3.43*** (.41)	−3.06*** (.53)	−1.31* (.77)
Model χ²	88.89***	5.18	16.36**	341.01***	37.43***	13.63**
n	3,157	1,054	2,103	3,159	1,055	2,104

Note. Standard errors are presented in parentheses.

p < .05. **p < .01. ***p < .001, one-tailed test.

According to Table 8, an androgenized finger ratio on the right hand (i.e., a low r2D:4D ratio) is associated with significantly greater reckless driving for males and females. However, a low right finger ratio was not associated with illegal motor racing for either gender.

Table 8.

Negative Binomial Regression Model Using r2D:4D Digit Ratio, Sex, Age, and Honesty as Predictors of Two Measures of Endangerment Crime.

Predictors	Illegal motor racing			Reckless driving
Predictors	Total sample	Males	Females	Total sample	Males	Females
r2D:4D ratio	−.02 (.03)	−.05 (.04)	−.07 (.05)	−.15*** (.04)	−.13** (.05)	−.16*** (.05)
Sex	1.60*** (.07)	—	—	1.25*** (.07)	—	—
Age	.03* (.01)	.08*** (.02)	−.10** (.03)	.03*** (.01)	.07*** (.01)	−.04* (.02)
Honesty	−.04* (.02)	.01*** (.03)	.04 (.03)	.25*** (.03)	.22** (.04)	.31*** (.05)
Constant	−5.53*** (.60)	−2.44*** (.43)	.17 (.68)	−4.03*** (.35)	−3.34*** (.47)	−2.91*** (.63)
Model χ²	958.93***	48.62***	18.84***	488.00***	82.39***	70.54***
n	3,147	1,050	2,097	2,913	986	1,927

Note. Standard errors are presented in parentheses.

p < .05. **p < .01. ***p < .001, one-tailed test.

Table 9 summarizes findings for three consensual/victimless crimes. It shows that the r2D:4D ratio was inversely associated with illegal drug use and illegal gambling to significant degrees for males and females. In the case of illegal commerce, however, the ratio was significantly linked to illegal commerce in the case of females only.

Table 9.

Negative Binomial Regression Model Using r2D:4D Digit Ratio, Sex, Age, and Honesty as Predictors of Three Categories of Consensual Crime.

Predictors	Drug distribution			Illegal drug use			Illegal gambling
Predictors	Total	Males	Females	Total	Males	Females	Total	Males	Females
r2D:4D ratio	−.18*** (.05)	−.05 (.06)	−.30*** (.06)	−.15*** (.03)	−.16** (.05)	−.15*** (.04)	−.28*** (.06)	−.45*** (.07)	.07 (.10)
Male	.87*** (.09)	—	—	.47*** (.06)	—	—	2.05*** (.13)	—	—
Age	−.04* (.02)	−.06*** (.03)	−.02 (.02)	−.04*** (.01)	−.03** (.01)	−.04** (.01)	−.002 (.02)	.02 (.02)	−.07 (.05)
Honesty	.48*** (.05)	.41*** (.06)	.57*** (.08)	.52*** (.03)	.36*** (.04)	.68*** (.04)	.30*** (.05)	.35*** (.06)	.16* (.08)
Constant	−5.11*** (.59)	−3.45*** (.81)	−6.09*** (.87)	−4.16*** (.35)	−2.24*** (.49)	−5.65*** (.50)	−4.99*** (.61)	−3.43*** (.70)	−3.20** (1.35)
Model χ²	285.55***	72.17***	112.52***	550.40***	123.04***	389.40***	479.20***	95.82***	6.94
n	3,158	1,055	2,103	3,158	1,055	2,103	2,913	985	1,928

Note. Standard errors are presented in parentheses.

p < .05. **p < .01. ***p < .001, one-tailed test.

Turning to the other variables included in the statistical models for Tables 4 through 9, one can see that sex is strongly and consistently related to all 13 crime categories. By contrast, age was inconsistently related to the offense measures in terms of either direction or statistical significance. Interestingly, the honesty variable (where respondents finished the questionnaire by indicating how carefully they read the questions and honestly answered them) was also somewhat irregularly associated with criminal involvement, but for most offenses, those who reported being the most honest indicated greater involvement than those who were the least honest. The two main exceptions had to do with sexual assault and serious property damage, where honesty was associated inversely with involvement.

Discussion

One reason criminologists continue to exclude biological variables from their efforts to explain criminality is that no irrefutable evidence for such influences has yet presented itself. Even the evidence that sex differences in criminality appear to be universal could conceivably be explained in social terms. Fausto-Sterling (1992), for instance, stated that universal sex differences in behavior could have resulted from the fact that “the entire population of the world all evolved from a small progenitor stock, and [from this simple fact many sex differences were] faithfully passed down a thousand times over.” However, such a line of reasoning would be unable to account for findings in the present study. Here, a physical trait that is determined prenatally was found to correlate with criminal involvement many years later. While this physical trait (the 2D:4D finger ratio) is itself associated with gender (Caramaschi et al., 2012; Fink et al., 2006), the present study shows that even when gender is eliminated as a variable, many significant associations between 2D:4D and offending behavior remain. All these associations are as one would suspect in light of an androgen-influence hypothesis.

To our knowledge, this is the first study to investigate the possibility that a measure of prenatal androgen exposure is associated with criminality. The fact that this biological variable was able to predict several forms of criminality even within each gender eliminates the possibility that social factors are mediating the relationship. Instead, the findings support a biosocial criminological theory formulated roughly a decade ago that asserts that brain exposure to androgens is a major contributor to criminality (Ellis, 2003).

Regarding limitations, the measure of r2D:4D that was used in this study was derived from self-reports. According to Hönekopp and Watson (2010), calculating 2D:4D ratios using self-reported digit lengths are about two thirds as accurate as calculations based on physical measurements made by trained researchers. Assuming that the inaccuracies resulting from self-reported measurement are random rather than systematic, it is reasonable to infer that the present measures provide only minimal estimates of the true associations between r2D:4D and criminality.

An important caveat surrounding the use of finger length measures to estimate prenatal androgen exposure is that many factors aside from androgens affect digit growth (McIntyre et al., 2005; Voracek et al., 2007). Furthermore, the availability of androgens in the body’s extremities cannot be considered entirely reflective of androgen levels within the nervous system (McIntyre, 2006; Talarovicová, Krsková, & Blazeková, 2009), and it is obviously the nervous system that is most relevant to altering behavior patterns.

Another limitation of the present study involves the fact that self-reported offending is far from complete or accurate (Moffitt & Silva, 1988; Thornberry & Krohn, 2000). In all likelihood, the unreliability in r2D:4D plus the limitations of self-reported offending means that our findings provide a minimal estimate of the real extent to which prenatal brain exposure to androgens affects later involvement in delinquency and crime. Furthermore, one should keep in mind that the present study was limited to college students whose involvement in crime tends to be considerably less than for populations at large.

Strengths of the present study include the fact that it was based on large samples drawn from two quite different cultures. The overall robustness of the links between r2D:4D and most forms of criminality makes these relationships virtually impossible to explain as coincidental. If confirmed, criminologists in the future need to incorporate androgens into their research and theorizing efforts.

Footnotes

Acknowledgements

We thank the following persons for helping to recruit research participants for this study: Drew H. Bailey, David Geary, Richard D. Hartley, Richard Lippa, Emi Prihatin, David Puts, Malini Ratnasingam, Anthony Walsh, and Alan Widmayer. Comments by Martin Voracek and anonymous reviewers on earlier versions of this paper were very helpful in bringing this research paper to completion.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: An administrative grant from the University of Malaya to its Department of anthropology and sociology helped to fund data collection and data entry.

References

Agnew

(1995). Testing the leading crime theories: An alternative strategy focusing on motivational processes. Journal of Research in Crime & Delinquency, 32, 363-398.

Akers

R. L.

(2009). Social learning and social structure: A general theory of crime and deviance. New York, NY: Transaction Publishers.

Allaway

H. C.

Bloski

T. G.

Pierson

R. A.

Lujan

M. E.

(2009). Digit ratios (2D: 4D) determined by computer-assisted analysis are more reliable than those using physical measurements, photocopies, and printed scans. American Journal of Human Biology, 21, 365-370.

Anderson

(2012). Comparing risk taking and digit ratio (2D:4D) in offenders and non-offenders. Plymouth Student Scientist, 5, 105-120.

Arnold

A. P.

(2009). The organizational-activational hypothesis as the foundation for a unified theory of sexual differentiation of all mammalian tissues. Hormones and Behavior, 55, 570-578.

Arrigo

B. A.

Young

(1998). Theories of crime and crimes of theorists. Theory & Psychology, 8, 219-252.

Berenbaum

S. A.

Duck

S. C.

Bryk

(2000). Behavioral effects of prenatal versus postnatal androgen excess in children with 21-hydroxylase-deficient congenital adrenal hyperplasia. Journal of Clinical Endocrinology & Metabolism, 85, 727-733.

Blanchard

Lyons

(2010). An investigation into the relationship between digit length ratio (2D: 4D) and psychopathy. British Journal of Forensic Practice, 12, 23-31.

Booth

Granger

D. A.

Mazur

Kivlighan

K. T.

(2006). Testosterone and social behavior. Social Forces, 85, 167-191.

10.

Brabin

Roberts

S. A.

Farzaneh

Fairbrother

Kitchener

H. C.

(2008). The second to fourth digit ratio (2D: 4D) in women with and without human papillomavirus and cervical dysplasia. American Journal of Human Biology, 20, 337-341.

11.

Breedlove

S. M.

(2010). Minireview: Organizational hypothesis: Instances of the fingerpost. Endocrinology, 151, 4116-4122.

12.

Brown

W. M.

Hines

Fane

B. A.

Breedlove

S. M.

(2002). Masculinized finger-length patterns in human males and females with congenital adrenal hyperplasia. Hormones and Behavior, 42, 380-386.

13.

Campbell

Murcer

(1990). Causes of crime: Uncovering a lay model. Criminal Justice and Behavior, 17, 410-419.

14.

Caramaschi

Booij

Petitcherc

Boivin

Tremblay

R. E.

(2012). Genetic and environmental contributions to saliva testosterone levels in male and female infant twins. Psychoneuroendocrinology, 37, 1954-1959.

15.

Cooper

J. A.

Walsh

Ellis

(2010). Is criminology moving toward a paradigm shift? Evidence from a survey of the American Society of Criminology. Journal of Criminal Justice Education, 21, 332-347.

16.

DiPrete

T. A.

Buchmann

(2006). Gender-specific trends in the values of education and the emerging gender gap in college completion. Demography, 43, 1-24.

17.

Ebeling

P. R.

(2010). Androgens and osteoporosis. Current Opinion in Endocrinology, Diabetes and Obesity, 17, 284-292.

18.

Eising

C. M.

Eikenaar

Schwabl

Groothuis

T. G. G.

(2001). Maternal androgens in black-headed gull (Larus ridibundus) eggs: Consequences for chick development. Proceedings of the Royal Society of London. Series B: Biological Sciences, 268, 839-846.

19.

Ellis

(2003). Genes, criminality, and evolutionary neuroandrogenic theory. In Walsh

Ellis

(Eds.), Biosocial criminology, challenging environmentalism’s supremacy (pp. 13-34). New York, NY: Nova Science.

20.

Ellis

(2005). Theoretically explaining biological correlates of criminal behavior. European Journal of Criminology, 2, 287-315.

21.

Ellis

(2011). Evolutionary neuroandrogenic theory and universal gender differences in cognition and behavior. Sex Roles, 64, 707-722.

22.

Ellis

Beaver

Wright

(2009). Handbook of crime correlates. Amsterdam, Netherlands: Academic Press.

23.

Erskine

(1974). Polls: Causes of crime. Public Opinion Quarterly, 38, 288-297.

24.

Falter

C. M.

Plaisted

K. C.

Davis

(2008). Visuo-spatial processing in autism—Testing the predictions of extreme male brain theory. Journal of Autism and Developmental Disorders, 38, 507-515.

25.

Fausto-Sterling

(1992). Myths of gender: Biological theories about women and men. New York, NY: Basic Books.

26.

Fink

Thanzami

Seydel

Manning

J. T.

(2006). Digit ratio and hand-grip strength in German and Mizos men: Cross-cultural evidence for an organizing effect of prenatal testosterone on strength. American Journal of Human Biology, 18, 776-782.

27.

Firebaugh

Dorius

S. F.

(2010). Trends in global gender inequality. Social Forces, 88, 1941-1968.

28.

Furnham

Henderson

(1983). Lay theories of delinquency. European Journal of Social Psychology, 13, 107-120.

29.

Garn

S. M.

Burdi

A. R.

Babler

W. J.

Stinson

(1975). Early prenatal attainment of adult metacarpal-phalangeal rankings and proportions. American Journal of Physical Anthropology, 43, 327-332.

30.

Ghiglieri

M. P.

(2000). The dark side of man: Tracing the origins of male violence. New York, NY: Basic Books.

31.

Gottfredson

M. R.

Hirschi

(1990). A general theory of crime. Stanford, CA: Stanford University Press.

32.

Hanoch

Gummerum

Rolison

(2012). Second-to-fourth digit ratio and impulsivity: A comparison between offenders and nonoffenders. PLoS ONE, 7, e47140.

33.

Harris

J. A.

Vernon

P. A.

Boomsma

D. I.

(1998). The heritability of testosterone: A study of Dutch adolescent twins and their parents. Behavior Genetics, 28, 165-171.

34.

Hilbe

(2011). Negative binomial regression (2nd ed.). New York, NY: Cambridge University Press.

35.

Hönekopp

(2011). Relationships between digit ratio 2D:4D and self-reported aggression and risk taking in an online study. Personality and Individual Differences, 51, 77-80.

36.

Hönekopp

Manning

Müller

(2006). Digit ratio (2D: 4D) and physical fitness in males and females: Evidence for effects of prenatal androgens on sexually selected traits. Hormones and Behavior, 49, 545-549.

37.

Hönekopp

Voracek

Manning

J. T.

(2006). 2nd to 4th digit ratio (2D: 4D) and number of sex partners: Evidence for effects of prenatal testosterone in men. Psychoneuroendocrinology, 31, 30-37.

38.

Hönekopp

Watson

(2010). Meta-analysis of digit ratio 2D:4D shows greater sex difference in the right hand. American Journal of Human Biology, 22, 619-630.

39.

Kemper

C. J.

Schwerdtfeger

(2009). Comparing indirect methods of digit ratio (2D: 4D) measurement. American Journal of Human Biology, 21, 188-191.

40.

Kimura

(1992). Sex differences in the brain. Scientific American, 267, 118-125.

41.

Kremelberg

(2011). Practical statistics. Thousand Oaks: CA: Sage.

42.

Kuijper

Lambalk

Boomsma

Van Der Sluis

Blankenstein

de Geus

Posthuma

(2007). Heritability of reproductive hormones in adult male twins. Human Reproduction, 22, 2153-2159.

43.

Lutchmaya

Baron-Cohen

Raggatt

Knickmeyer

Manning

(2004). 2nd to 4th digit ratios, fetal testosterone and estradiol. Early Human Development, 77, 23-28.

44.

Malas

M. A.

Dogan

Hilal Evcil

Desdicioglu

(2006). Fetal development of the hand, digits, and digit ratio (2D:4D). Early Human Development, 82, 469-475.

45.

Manning

J. T.

Scutt

Wilson

Lewis-Jones

D. I.

(1998). The ratio of 2nd to 4th digit length: A predictor of sperm numbers and concentrations of testosterone, luteinizing hormone and estrogen. Human Reproduction, 13, 3000-3004.

46.

Manning

J. T.

Stewart

Bundred

P. E.

Trivers

R. L.

(2004). Sex and ethnic differences in 2nd to 4th digit ratio of children. Early Human Development, 80, 161-168.

47.

McIntyre

M. H.

(2006). The use of digit ratios as markers for perinatal androgen action. Reproductive Biology and Endocrinology, 4, 1-9.

48.

McIntyre

M. H.

Ellison

P. T.

Lieberman

D. E.

Demerath

Towne

(2005). The development of sex differences in digital formula from infancy in the Fels Longitudinal Study. Proceedings of the Royal Society B: Biological Sciences, 272, 1473-1479.

49.

Moffitt

T. E.

Silva

P. A.

(1988). Self-reported delinquency: Results from an instrument for New Zealand. Australian & New Zealand Journal of Criminology, 21, 227-240.

50.

Muller

D. C.

Giles

G. G.

Bassett

Morris

H. A.

Manning

J. T.

Hopper

J. L.

. . .Severi

(2011). Second to fourth digit ratio (2D:4D) and concentrations of circulating sex hormones in adulthood. Reproductive Biology and Endocrinology, 9, 1-12.

51.

Pinna

Costa

Guidotti

(2005). Changes in brain testosterone and allopregnanolone biosynthesis elicit aggressive behavior. Proceedings of the National Academy of Sciences of the United States of America, 102, 2135-2141.

52.

Reinisch

J. M.

(1974). Fetal hormones, the brain, and human sex differences: An heuristic, integrative review of the recent literature. Archives of Sexual Behavior, 3, 51-90.

53.

Rogers

L. J.

(2002). Sexing the brain. New York, NY: Columbia University Press.

54.

Romeo

(2003). Puberty: A period of both organizational and activational effects of steroid hormones on neurobehavioural development. Journal of Neuroendocrinology, 15, 1185-1192.

55.

Rubinow

D. R.

Schmidt

P. J.

(1996). Androgens, brain, and behavior. American Journal of Psychiatry, 153, 974-984.

56.

Schwerdtfeger

Heims

Heer

(2010). Digit ratio (2D:4D) is associated with traffic violations for male frequent car drivers. Accident Analysis & Prevention, 42, 269-274.

57.

Sharpe

R. M.

(1998). The roles of estrogen in the male. Trends in Endocrinology & Metabolism, 9, 371-377.

58.

Simon

N. G.

(2006). Androgens and aggression. In Nelson

R.J.

(Ed.), Biology of aggression (pp. 211-230). New York, NY: Oxford University Press.

59.

Sims

N. A.

Brennan

Spaliviero

Handelsman

D. J.

Seibel

M. J.

(2006). Perinatal testosterone surge is required for normal adult bone size but not for normal bone remodeling. American Journal of Physiology-Endocrinology and Metabolism, 290, E456-E462.

60.

Stalans

L. J.

Lurigio

A. J.

(1990). Lay and professionals’ beliefs about crime and criminal sentencing: A need for theory, perhaps schema theory. Criminal Justice and Behavior, 17, 333-349.

61.

Talarovicová

Krsková

Blazeková

(2009). Testosterone enhancement during pregnancy influences the 2D:4D ratio and open-field motor activity of rat siblings in adulthood. Hormones and Behavior, 55, 235-239.

62.

Thornberry

T. P.

Krohn

M. D.

(2000). The self-report method for measuring delinquency and crime. Criminal Justice, 4, 33-83.

63.

Udry

J. R.

(1990). Biosocial models of adolescent problem behaviors. Biodemography and Social Biology, 37, 1-10.

64.

Volavka

(2002). Neurobiology of violence. Washington, DC: American Psychiatric Publishing.

65.

vom Saal

F. S.

(1983). Models of early hormonal effects on intrasex aggression in mice. In Svare

(Ed.), Hormones and aggressive behavior (pp. 197-222). New York, NY: Plenum.

66.

Voracek

Manning

J. T.

Dressler

S. G.

(2007). Repeatability and inter-observer error of digit ratio (2D: 4D) measurements made by experts. American Journal of Human Biology, 19, 142-146.

67.

Williams

T. J.

Pepitone

M. E.

Christensen

S. E.

Cooke

B. M.

Huberman

A. D.

Breedlove

N. J.

. . .Breedlove

(2000). Finger-length ratios and sexual orientation. Nature, 404, 455-456.

68.

Wright

J. P.

Beaver

K. M.

DeLisi

Vaughn

M. G.

Boisvert

Vaske

(2008). Lombroso’s legacy: The miseducation of criminologists. Journal of Criminal Justice Education, 19, 325-338.