REVISITING THE CRIMINOLOGICAL CONSEQUENCES OF EXPOSURE TO FETAL TESTOSTERONE: A META‐ANALYSIS OF THE 2D:4D DIGIT RATIO*

• DOI: http://doi.org/10.1111/1745-9125.12115
• Author: TRAVIS C. PRATT,FRANCIS T. CULLEN,JILLIAN J. TURANOVIC
• Date: 01 November 2016
• Published date: 01 November 2016

REVISITING THE CRIMINOLOGICAL

CONSEQUENCES OF EXPOSURE TO FETAL

TESTOSTERONE: A META-ANALYSIS OF THE 2D:4D

DIGIT RATIO∗

TRAVIS C. PRATT,1JILLIAN J. TURANOVIC,2

and FRANCIS T. CULLEN3

1Corrections Institute, University of Cincinnati

2College of Criminology and Criminal Justice, Florida State University

3School of Criminal Justice, University of Cincinnati

KEYWORDS: biosocial criminology, fetal testosterone, 2D:4D ratio, crime, aggression,

impulsivity

As criminology has become more interdisciplinary in recent years, biosocial crim-

inology has earned a place at the table. Although this perspective comes in many

forms, one important proposition has gained increasing attention: that the 2D:4D fin-

ger digit ratio—a purported physical biomarker for exposure to fetal testosterone—is

related to criminal, aggressive, and risky/impulsive behavior. Strong claims in the lit-

erature have been made for this link even though the findings seem to be inconsistent.

To establish the empirical status of this relationship, we subjected this body of work

to a meta-analysis. Our multilevel analyses of 660 effect size estimates drawn from

47 studies (14,244 individual cases) indicate a small overall effect size (mean r =.047).

Moderator analyses indicate that this effect is rather “general” across methodological

specifications—findings that are at odds with theoretical propositions that specify the

importance of exposure to fetal testosterone in predicting criminal and analogous be-

havior later in life. We conclude with a call for exercising caution over embracing the

findings from one or two studies and instead highlight the importance of systematically

organizing the full body of literature on a topic before making decisions about what

does, and what does not, predict criminal and analogous behavior.

For most of its intellectual life, criminology was “owned” by sociology (Laub, 2004).

During this time, those who thought that other academic disciplines might have some-

thing important to say about criminal behavior were either shoved to the margins of the

field or were tarred and feathered for pushing ideas that could be socially dangerous if

placed in the wrong hands (e.g., Glueck and Glueck, 1950; Hirschi and Hindelang, 1977;

see the discussion by Laub and Sampson, 1991). Yet in recent years, criminology has

∗Additional supporting information can be found in the listing for this article in the Wiley Online

Library at http://onlinelibrary.wiley.com/doi/10.1111/crim.2016.54.issue-4/issuetoc.

The authors would like to thank the anonymous reviewers for their helpful comments and Wayne

Osgood for his guidance throughout the review process.

Direct correspondence to Travis C. Pratt, University of Cincinnati Corrections Institute, 2840

Bearcat Way, Cincinnati, OH 45221 (e-mail: pratttc@ucmail.uc.edu).

C2016 American Society of Criminology doi: 10.1111/1745-9125.12115

CRIMINOLOGY Volume 54 Number 4 587–620 2016 587

588 PRATT, TURANOVIC & CULLEN

become much more inclusive and interdisciplinary, due in no small part to the infusion

of ideas gleaned from fields like developmental psychology (Farrington, 1995; Moffitt,

1993; Steinberg and Morris, 2001), behavioral economics (Anwar and Loughran, 2011;

Kahneman, 2011; Pogarsky and Piquero, 2003), and public health (Finkelhor, 2008; Wolf,

Gray, and Fazel, 2014). In the process, what is now called “biosocial criminology” has

found itself with a place at the criminological table (Walsh and Beaver, 2009). And

although it is not without its critics (Carrier and Walby, 2014), the accumulated evi-

dence shows clearly that this perspective has considerable intellectual value (Barnes and

Boutwell, 2015; Wright and Boisvert, 2009). In short, biosocial criminology is here to stay,

and for good reason.

There are, however, some heated disagreements within the biosocial criminology camp

concerning the best way to go about studying criminal and other risky behaviors from

a biosocial standpoint (e.g., the recent exchange between Burt and Simons, 2014, and

Barnes et al., 2014). This debate is complex and multidimensional, with disagreements

ranging from how genes and the environment interact (Burt and Simons, 2015) to how

well we can meet certain statistical assumptions (Wright et al., 2015). But at the end of

the day, much of the difference of opinion arguably centers around the relative utility—

or necessity—of using direct indicators (e.g., biological/genetic markers; see Beaver et al.,

2007; Caspi et al., 2010; Simons et al., 2011) versus indirect approaches (e.g., heritability

estimates from twin studies; see Barnes, Beaver, and Boutwell, 2011; Beaver et al., 2009;

Moffitt and Beckley, 2015) when studying the biosocial processes underlying criminal

behavior.

A similar debate is also bubbling up around one of the latest purported indirect indica-

tors of biological processes: the 2D:4D digit ratio. The term “2D” refers to the length of

the index finger, whereas the term “4D” refers to the length of the ring finger. The ratio

is calculated by dividing the length of the index finger, which is shorter, by the length of

the ring finger, which is longer. This ratio is seen as a physical biomarker for how much

fetal testosterone a person was exposed to (Manning et al., 2003; see also McIntyre et al.,

2009; Talarovicova, Krskova, and Blazekova, 2009), which is assumed to be related to ag-

gressive, risky/impulsive, and even criminal behavior later in life (Ellis, 2005; Williams,

Greenhalgh, and Manning, 2003). Although the results of the empirical studies that have

addressed the digit ratio-criminal/analogous behavior link have been mixed (e.g., com-

pare Anderson, 2012; Kim, Kim, and Kim, 2014; Stenstrom et al., 2011), a recent study

by Hoskin and Ellis (2015) found extremely large and robust effect sizes—values rising

up to .40 and above—for the digit ratio and criminal behaviors ranging from serious as-

sault to reckless driving to fraud. These effects are about twice as large as we typically see

for variables like low self-control (de Ridder et al., 2012; Pratt and Cullen, 2000), deviant

peer influences (Andrews and Bonta, 2010; Nguyen and McGloin, 2013; Pratt et al., 2010),

and collective efficacy (Sampson, 2006)—factors that count themselves among the core of

criminology (Cullen, Wright, and Blevins, 2006).

For criminologists, the study by Hoskin and Ellis (2015) is of particular importance. Be-

cause of its publication in the American Society of Criminology’s flagship journal, Crim-

inology, the article presented fascinating findings that may have the potential to exert

considerable influence on the field. We do not wish to critique Hoskin and Ellis’s study

or the results that they found, but we instead want to caution against affording any sin-

gle study too much weight. In this case, because virtually all of the research on the be-

haviors correlated with the 2D:4D digit ratio has been published in journals outside of

EXPOSURE TO FETAL TESTOSTERONE REVISITED 589

criminology (cf. Ellis and Hoskin, 2015a), it is unlikely to be read by most criminologists.

The temptation to rely on the most readily available work in a prominent disciplinary

journal is therefore high. There is, however, a growing body of research, across a wide ar-

ray of scientific and social scientific fields, showing that the findings of individual studies—

even those of high methodological quality—do not always replicate well (e.g., Ioannidis,

2005a, 2005b; Lehrer, 2010; Open Science Collaboration, 2015). Thus, establishing confi-

dence that empirical relationships really exist first requires the systematic assessment of

all of the studies on the topic at hand (Schmidt and Oh, 2014).

Accordingly, the purpose of the present study is to subject this body of literature to

a meta-analysis. Doing so is particularly useful when the results from a body of empiri-

cal literature on a common research question are inconsistent (Pratt, 2010). Our analy-

sis is based on 660 effect size estimates drawn from 47 empirical studies of the relation-

ship between the 2D:4D digit ratio and crime and crime-analogous (e.g., aggression and

risky/impulsive) behavior, which we analyze with the appropriate multilevel modeling

techniques. In doing so, two interrelated objectives serve as our guides. First, we seek to

establish the overall effect size of the 2D:4D digit ratio to these criminological outcomes

to determine whether this effect is large enough that it should demand serious attention

from criminologists to explain its existence. Second, our analyses are aimed at uncover-

ing how the magnitude of these effects may vary according to, or are robust across, the

methodological choices made by the scholars who have produced this work (e.g., across

different types of samples and kinds of outcomes examined). These analyses are intended

to inform the field about the extent to which the 2D:4D digit ratio is—or is not—likely to

be a useful and productive addition to criminological research in the future. And yet in

the end, our broader purpose is to shed light on how devoting greater attention to mea-

surement, replication, and the systematic organization of knowledge will be critical to the

scientific legitimacy of criminology as the field moves forward.

CRIMINOLOGY, BIOSOCIAL PERSPECTIVES, AND THE 2D:4D

RATIO

Biological explanations of human behavior can be scary. Beginning in the early 1900s,

for example, these ideas legitimized Nazism, racism, sexism, and other “isms” that we

find ideologically disquieting (e.g., the discussions by Heylen et al., 2015, and Rowe and

Osgood, 1984). In response, criminology not only rejected biological explanations of crim-

inal behavior, but also rejected individualistic explanations altogether in favor of perspec-

tives that emphasized factors like structural disadvantage and the cultural responses to it

(Merton, 1938; Shaw and McKay, 1942; Sutherland, 1947). Things stayed that way for a

long time, but the winds have shifted in recent years as interdisciplinary research demon-

strating that humans are not “blank slates” at birth has become too convincing to ignore

(Pinker, 2003). Of course, shaking off the Lombrosian past has proven to be an uphill

battle (Wright et al., 2008), but a large body of work has emerged in recent years that

has given biosocial criminology a hard-earned level of empirical legitimacy (Rafter, 2008;

Wright and Cullen, 2012).

And as it has developed, contemporary biosocial criminology now comes in many forms

and covers a considerable amount of intellectual territory (for a thorough review, see

Barnes, Boutwell, and Beaver, 2015). Yet when it comes to thinking about the poten-

tial criminogenic effects of factors such as exposure to fetal testosterone, it is important