Artículo en PDF
How to cite
Complete issue
More information about this article
Journal's homepage in redalyc.org
Sistema de Información Científica
Red de Revistas Científicas de América Latina y el Caribe, España y Portugal
The rediscovery of sex differences (men < women) in the length
ratio of the index finger (2D) to the ring finger (4D) (2D:4D;
Manning, Scutt, Wilson, & Lewis-Jones, 1998), along with a
hypothesis about the origin of these, has spurred ongoing research
interest in psychology and adjacent disciplines. Because this sex
effect is determined in utero and 2D:4D shows little plasticity
thereafter, 2D:4D has been suggested as an indirect marker for the
organizational (permanent) masculinizing effects of prenatal
androgens on the brain, behavior, and physique (Manning, 2002).
Given this hypothesis is correct, it would straightforwardly
allow for comfortable (non-invasive) tests of the biological bases
of behavior, cognition, personality, and other individual difference
variables that show sex differences and therefore conceivably are
partly influenced by prenatal sex-hormone action. Unsurprisingly,
this idea has attracted wide interest, resulting in a sizeable
literature now approaching 250 reports (partial reviews: Manning,
2002; Putz, Gaulin, Sporter, & McBurney, 2004).
2D:4D research also has a number of problems, including usually
modest effects, preponderantly unreplicated findings, statistically
underpowered (and thus type I error-prone) studies reporting positive
findings, a spate of recently published nonreplications, and single-
sample studies (without replication attempts) still being the norm.
Replicated nil associations of digit ratio (2D:4D) and absolute finger
lengths with implicit and explicit measures of aggression
Martin Voracek and Stefan Stieger
University of Vienna (Austria)
Preliminary evidence suggests that within-sex individual variation in the length ratio of the index finger
to the ring finger (2D:4D), a putative marker of prenatal androgen levels, may be more strongly or
consistently related to implicit measures (Implicit Association Test, IAT) than to corresponding explicit
(self-report) measures of candidate personality traits. The underlying logic is that the non-introspective
IAT may reflect earlier and inaccessible experiences, whereas introspective self-reports reflect more
recent and necessarily accessible experiences. Associations of 2D:4D and absolute finger length (a
marker of pubertal-adolescent androgen levels) with implicit versus explicit aggression measures (IAT
vs. Buss-Perry Aggression Questionnaire and a feeling thermometer) were examined in two samples
with identical procedures (
Ns
= 244 and 233). Attesting to procedural validity and data typicality,
several experiments of related research were replicated in both samples, including sex differences in
2D:4D and theory compliant interrelations among psychometric measures. However, no theory
compliant, reliable, or replicable associations of 2D:4D or finger length with implicit or explicit
aggression measures resulted. These nil findings cast doubt on hypothesized advantages of implicitly
(over explicitly) measured target traits for 2D:4D research. More generally, they add to a growing
number of replication failures in this area.
Replicación de asociaciones nulas del ratio del segundo y cuarto dedo (2D:4D) y la longitud absolu-
ta de los dedos con medidas de agresión implícitas y explícitas.
Evidencias preliminares sugieren que
la variabilidad intersexual en la relación entre la longitud del 2º dedo (índice) y del cuarto dedo
(2D:4D), como marcador indirecto de los niveles de andrógenos prenatales, puede poseer una fuerte y
coherente relación con medidas implícitas (Test de Asociación Implícita, IAT) que corresponden a me-
didas explícitas (autoinforme) de rasgos de la personalidad. La lógica fundamental es que el IAT-no in-
trospectivo puede reflejar las experiencias más tempranas e inaccesibles, mientras que los autoinfor-
mes reflejan las experiencias más recientes y necesariamente accesibles. La asociación entre el 2D:4D
y la longitud absoluta del dedo (como marcador de los niveles de andrógenos en la etapa de la puber-
tad-adolescencia) con las medidas de agresión implícitas vs las explícitas (IAT vs. Cuestionario de
Agresión de Buss-Perry y termómetro de sentimiento) fueron analizadas mediante los mismos proce-
dimientos en dos muestras (
Ns
= 244 and 233). Para verificar la validez del procedimiento y la norma-
lidad de los datos, varias experiencias de la investigación fueron replicadas en ambas muestras, inclu-
yendo las diferencias sexuales en el 2D:4D y las de la interrelación de la teoría conformista entre
medidas psicométricas.
Fecha recepción: 1-10-08 • Fecha aceptación: 11-1-09
Correspondencia: Martin Voracek
Dept. of Basic Psychological Research - School of Psychology
University of Vienna - Liebiggasse, 5, Rm 03-46
A-1010 Vienna (Austria)
E-mail: martin.voracek@univie.ac.at
Psicothema 2009. Vol. 21, nº 3, pp. 382-389
ISSN 0214 - 9915 CODEN PSOTEG
www.psicothema.com
Copyright © 2009 Psicothema
Pág. 382-
29/6/09
19:40
Página 382
This possibility seems not remote. Publication bias was clearly
demonstrated for fluctuating asymmetry research (Palmer, 2000),
a field sharing several features with 2D:4D research (e.g., small
effects, measurement and methodology issues, underpowered
studies, replication failures, rapid growth of the literature, a few
productive labs, and competitive publishing).
To give one example, an initial report (Csathó, Osváth, Bicsák,
Karádi, Manning, & Kállai, 2003), based on merely 46 female
Hungarian students, suggested associations of 2D:4D with sex-role
orientation. According to ISI Web of Knowledge, this report ranks
among the top 15 of most frequently cited papers of 2D:4D research.
However, five subsequent reports, including large-scale and multi-
sample evidence (Hampson, Ellis, & Tenk, 2008; Lippa, 2006;
Rammsayer & Troche, 2007; Schmukle, Liesenfeld, Back, & Egloff,
2007; Troche, Weber, Hennigs, Andresen, & Rammsayer, 2007),
and at least six unpublished theses (Evardone, 2006; Fawcett, 2003;
Meingaßner, 2003; Robinson, 2005; Schicker, 2005; Vonnahme,
2005) have failed to replicate such associations. Two further studies
with positive findings (Beech & Mackintosh, 2005; Scarbrough &
Johnston, 2005) were based on similarly small samples (
N
s= 60 and
41). On the whole, there is more evidence against associations of
2D:4D with sex-role orientation than support for these.
In this context, one study (Schmukle et al., 2007) is of particular
importance, because it was hypothesized that implicit (automatic or
nonconscious) measures, as opposed to explicit (controlled or
conscious) measures, might entertain more reliable associations
with 2D:4D. This is because explicit mental representations, as
gauged by self-report measures, exploit propositional structures
that are accessible through introspection, malleable, and thus
reflect more recent events/experiences, whereas implicit mental
representations build on associative structures that are unaccessible
through introspection, less malleable, and thus may reflect earlier
events/experiences (Fazio & Olson, 2003).
Using the Implicit Association Test (IAT; Greenwald, McGhee,
& Schwartz, 1998), a widely applied experimental method for
measuring implicit social cognition (Lane, Banaji, Nosek, &
Greenwald, 2007), Schmukle et al. (2007) indeed obtained
stronger associations (significant for men, but not significant for
women) of 2D:4D with sex-role orientation, when measured
implicitly rather than explicitly. However, this evidence again was
based on a small (39 men and 42 women) single-sample design
without replication attempt.
There are several generic approaches for clarifying conflicting
research literatures, most of which have already been tried within
2D:4D research, including meta-analyses (e.g., Puts, McDaniel,
Jordan, & Breedlove, 2008), individual-data reanalyses of studies
(e.g., McFadden, Loehlin, Breedlove, Lippa, Manning, & Rahman,
2005), and multi-sample approaches (e.g., Troche et al., 2007).
Replicability of findings, a straightforward scientific criterion, remains
the
via regis
for establishing and furthering knowledge (Hendrick,
1990). For clarity, replications are differentiated into exact (or direct)
vs. systematic (or conceptual) ones. Exact replications imply repeating
studies using identical procedures, thus supporting (if successful) or
contesting (if unsuccessful) a theory. Systematic replications use
theoretical, procedural, or operational changes to the original study,
thus extending (if successful) or limiting (if unsuccessful) a theory.
In the present research, we applied both types of replications.
The rationale was as follows: if candidate individual difference
variables indeed show stronger or more consistent associations
with 2D:4D when measured implicitly rather than explicitly
(Schmukle et al., 2007), this should generalize to other traits than
sex-role orientation. Several studies have investigated associations
of 2D:4D with types of sexually differentiated aggression (Austin,
Manning, McInroy, & Mathews, 2002; Bailey & Hurd, 2005;
Benderlioglu & Nelson, 2004; Coyne, Manning, Ringer, & Bailey,
2007; Gallup, White, & Gallup, 2007; Hampson et al., 2008;
Kuepper & Hennig, 2007; McIntyre, Barrett, McDermott,
Johnson, Cowden, & Rosen, 2007; Millet & Dewitte, 2007). Of
note, their findings appear less conflicting than the 2D:4D
literature about sex-role orientation reviewed above and thus this
line of inquiry possibly is more fruitful. A detailed review of these
accounts is beyond the intended scope of the present work, but,
suffice it to say, that the majority of the studies cited above yielded
theory compliant within-sex associations of 2D:4D with various
measures of aggression. Also, there is evidence for an early
emergence of sex differences in aggression and related traits (e.g.,
childhood play) that are influenced by differences in early
exposure to androgens (Hines, 2008). We therefore selected
aggression (measured explicitly vs. implicitly) as the target trait
for investigating associations with 2D:4D. This attempt of
demonstrating generalizability constituted systematic replication.
As a further procedural extension, we considered two
additional measures of hand anatomy that similarly are thought to
be putative pointers to sex-hormonal exposure. First, D
R-L
, the
directional asymmetry (right-minus-left difference) in 2D:4D,
typically is lower in men than in women (Manning, Churchill, &
Peters, 2007; Voracek, Dressler, & Manning, 2007) and has
therefore been proposed as another putative marker for prenatal
androgen levels (Manning, 2002, pp. 21-22). For example, Coyne
et al. (2007) found D
R-L
to be negatively correlated with indirect
aggression in women. Second, absolute finger length (2D and 4D),
for which within-sex correlations with adult height are only
moderately positive (merely hovering around .50; Lippa, 2003),
seems to reflect pubertal-adolescent androgen levels, because the
sex difference therein is negligible prior to puberty, but very large
thereafter (Lippa, 2006). Because of the only loose association
with adult height and further because of evidence for significant
cross-talk of the androgen receptor (the target of testosterone) with
growth-factor signalling pathways (Nieschlag, Behre, &
Nieschlag, 2004), measures like absolute finger length cannot be
regarded as pointers to growth-hormone exposure exclusively, but
may well be appropriate markers for pubertal-adolescent
testosterone exposure. Up to now, the measure of absolute finger
length has only rarely been considered in 2D:4D research (e.g.,
Jackson, 2008).
We refrained from basing conclusions on findings yielded from
a single sample, even when this was considerably larger than the
one of the only predecessor study (Schmukle et al., 2007). We
therefore collected data from a second, independent sample, using
identical procedures. That way, we conducted an exact replication
of our own research.
Methods
Participants
Sample I were 114 men and 130 women, ages 16 to 59 years
(
M
= 27.5,
SD
= 10.3). Sample II were 101 men and 132 women,
REPLICATED NIL ASSOCIATIONS OF DIGIT RATIO (2D:4D) AND ABSOLUTE FINGER LENGTHS WITH IMPLICIT AND EXPLICIT MEASURES OF AGGRESSION
383
Pág. 382-
29/6/09
19:40
Página 383
population who volunteered for this research.
Instruments
Explicit aggression measures
. The German form (Herzberg,
2003) of the most widely used self-report measure of aggression
(Buss-Perry Aggression Questionnaire, BPAQ; Buss & Perry,
1992) was administered. This assesses the trait facets Physical
aggression, Verbal aggression, Anger, and Hostility (8, 5, 6, and 8
items) on 5 point scales (-2:
extremely uncharacteristic of me
, +2:
extremely characteristic of me
). Following standard practice in
IAT research (Greenwald et al., 1998), aggression was also
explicitly assessed with a feeling thermometer (0 to 100 degrees).
Implicit aggression measure
. IATs are now widely used
keyboard-based rapid sorting tasks of stimuli presented on
computer screens, whereby response latencies are analyzed to
indirectly measure the relative strengths of automatic,
nonconscious associations between paired concepts (commonly,
between categories
Pleasant
vs.
Unpleasant
, coupled with two
distinctive attributes). Sorting is easier (faster and less error-prone)
when two concepts sharing a response key are strongly
automatically associated than when less so (Greenwald et al.,
1998). In the traditional IAT, genuinely intrapersonal associations
are contaminated with extrapersonal (cultural) ones, because
categories
Pleasant
vs.
Unpleasant
are likely perceived as
normative. To reduce this contamination, personalized IAT
variants (Olson & Fazio, 2004) substitute these with idiosyncratic
categories (
I like
vs.
I don’t like
) and use evaluation-laden stimuli
with strong individual preferences, but no social consensus. A
considerable body of evidence attests to the predictive validity of
IAT measures using a variety of attitudes and self-concepts (for
reviews, see Fazio & Olson, 2003; Greenwald & Nosek, 2001).
A personalized aggression IAT was assembled as follows:
stimuli for the idiosyncratic categories (
I like
vs.
I don’t like
:
airplanes, beer, cleaning house, cigarettes, coffee, country music,
darkness, disco, eating, football, garlic, jogging, Monday,
motorcycles, opera, romance novels, spinach, storms, television,
tequila) were adopted (Olson & Fazio, 2004) and those for the
attributes (
Aggression
: anger, danger, fight, hothead, pain,
provocation, punch, revenge, violence, weapon;
Peace
: calmness,
freedom, health, heaven, joy, love, loyalty, pleasure, relaxation,
tenderness) selected by the authors.
For a schematic overview, see Table 1. Stimuli from the
categories (
I like
vs.
I don’t like
) and the attributes (
Aggression
vs.
Peace
) were presented in 5 blocks. Practice blocks (1, 2, and 4)
introduced concepts and stimuli (with each word presented twice).
Blocks 1 and 4 were similar, except that attribute position was
switched. Critically, concepts were differentially paired in Blocks
3 and 5, in order to measure their degree of association. Order of
these blocks was not counterbalanced to maximize individual
differences measurement and thus to enable unconstrained target-
trait correlations (Egloff & Schmukle, 2002).
The aggression IAT was run on laptops, with concept labels
presented in the right and left upper screen corners and stimuli
midway (intertrial interval: 150 ms). Participants were instructed
to react as quickly as possible and used keys «E» vs. «I» for
sorting to the left vs. right. Response latencies between stimulus
presentation and keystroke were recorded throughout Blocks 1 to
5 in ms.
Finger-length
measures
.
Following
published
2D:4D
ascertainment standards (Voracek, Manning, & Dressler, 2007),
flatbed-scanned images of participants’ right and left palms were
produced. From high-resolution laser printouts of scanned images,
length of 2D and 4D was measured triplicate, using the same image,
from the middle of the flexion crease proximal-most to the palm to
the fingertip by experienced, mutually blinded investigators (not the
authors), using digital vernier calipers accurate to 0.01 mm. Average
measurements were used to calculate right-hand and left-hand digit
ratio (R2D:4D and L2D:4D).
Procedure
After reporting basic demographic information, participants
completed the IAT, the BPAQ, and the feeling thermometer in
quiet facilities. Presenting an IAT prior to corresponding explicit
measures minimizes carry-over or priming effects (if any) among
measures, because implicit measures by their nature are less likely
to elicit such effects (Egloff & Schmukle, 2002). Finally, palm
scans were taken and participants thanked and debriefed.
Data analysis
Calculation of individual IAT results followed standard
practice, namely the originally proposed IAT effect algorithm (for
full procedural details, see Greenwald et al., 1998) which
continues to be used by a clear majority of IAT studies. Briefly, the
IAT effect (in units of logged milliseconds) is the difference
between the mean natural log transformed valid response latencies
from Blocks 5 and 3, whereby higher (positive) values index less
and lower (negative) values more aggression (i.e., compared with
the explicit aggression measures, the IAT was reversely scored).
Following Olson and Fazio (2004), participants with > 20% errors
in attribute trials were discarded from analysis (see Table 2 note;
by definition, response errors in idiosyncratic category trials are
undetectable). Also following standard practice (Greenwald et al.,
1998), calculation of the reliability of the IAT was based on the 78
response-time differences from the trials of Block 5 and 3 (the first
two out of the 80 trials of these blocks are omitted).
Interobserver repeatabilities of finger-length measurements
were assessed with single-score intraclass correlation coefficients
(
ICC
; two-way mixed-effects model with absolute-agreement
definition; Voracek et al., 2007). Significance was set to
p
<.05
(two-tailed), and
p
rep
, the probability of replicating an effect
(Killeen, 2005), was calculated for selected findings for
demonstration purposes.
MARTIN VORACEK AND STEFAN STIEGER
384
Table 1
IAT task sequence
Block
Trials
Block type
Concept label
Left key assignment
Right key assignment
1
40
Practice
Peace
Aggression
2
40
Practice
I like
I don’t like
3
80
Test
Peace
+
I like
Aggression
+
I don’t like
4
40
Practice
Aggression
Peace
5
80
Test
Aggression
+
I like
Peace
+
I don’t like
Pág. 382-
29/6/09
19:40
Página 384
Sex differences in variables
Descriptive statistics and tests for sex differences are shown in
Table 2. Replicating previous evidence (Manning, 2002) in both
samples, men had lower 2D:4D than women. This sex effect was
small-to-medium, somewhat weaker for L2D:4D than for R2D:4D,
and nearly absent for D
R-L
. Sex differences in absolute finger length
were large, but invariably larger for 4D than for 2D, which pattern
is the basis for the sex effect in 2D:4D.
Implicitly measured aggression was somewhat (but not
significantly) stronger in men than in women in both samples.
Explicitly measured aggression showed no sex effect (feeling
thermometer) or generally small ones that were inconsistent across
samples (BPAQ). Specifically, significantly higher Physical
aggression scores of men relative to women in Sample I failed to
replicate in Sample II, and significant sex differences in Anger
Reliabilities of variables
These are displayed in Table 3 (diagonal entries). Consistent with
prior evidence, in both samples finger-length and digit ratio
measurements were highly repeatable (Voracek et al., 2007), internal
consistency figures of the IAT high (Lane et al., 2007), and those of
BPAQ subscales tallied to reference values (Herzberg, 2003).
Sex-specific associations among variables
All interrelations of variables are set out in Table 3 (off-
diagonal entries). As expected, correlations among absolute finger
lengths and digit ratios were strongly positive (upper-left corner
entries) in both samples. Also, correlations among BPAQ
REPLICATED NIL ASSOCIATIONS OF DIGIT RATIO (2D:4D) AND ABSOLUTE FINGER LENGTHS WITH IMPLICIT AND EXPLICIT MEASURES OF AGGRESSION
385
Table 2
Sex differences in variables under study
Variable
Men
M
(
SD
)
Women
M
(
SD
)
tpd
Digit ratio
R2D:4D
0.957 (0.032)
0.969 (0.032)
-3.02
.003
-0.37
0.958 (0.029)
0.968 (0.029)
-2.70
.007
-0.34
L2D:4D
0.956 (0.032)
0.967 (0.029)
-2.67
.008
-0.36
0.956 (0.026)
0.964 (0.031)
-2.16
.03
-0.28
D
R-L
0.001 (0.019)
0.003 (0.026)
-0.65
.52
-0.09
0.002 (0.027)
0.004 (0.031)
-0.51
.61
-0.07
Absolute finger length (mm)
R2D
74.20 (3.99)
68.39 (4.43)
10.69
<.001
1.37
72.43 (4.04)
67.55 (4.44)
8.64
<.001
1.14
R4D
77.59 (4.43)
70.61 (4.89)
11.66
<.001
1.49
75.64 (3.96)
69.81 (4.71)
10.02
<.001
1.32
L2D
74.15 (4.33)
67.82 (4.52)
11.14
<.001
1.43
72.59 (4.55)
67.09 (4.26)
9.48
<.001
1.25
L4D
77.59 (4.58)
70.20 (4.82)
12.24
<.001
1.57
75.95 (4.73)
69.63 (4.86)
9.96
<.001
1.32
Average finger length
75.88 (4.01)
69.25 (4.49)
12.09
<.001
1.56
74.15 (3.97)
68.52 (4.17)
10.44
<.001
1.38
Aggression measures
IAT (log ms)
0.26 (0.22)
0.31 (0.24)
-1.57
.12
-0.22
0.27 (0.24)
0.30 (0.22)
-1.04
.30
-0.13
BPAQ-Physical aggression
-0.98 (0.73)
-1.34 (0.57)
4.27
<.001
0.55
-0.31 (0.51)
-0.29 (0.55)
-0.19
.85
-0.04
BPAQ-Verbal aggression
0.03 (0.64)
0.05 (0.60)
-0.22
.82
-0.03
-0.37 (0.66)
-0.52 (0.63)
1.66
.10
0.23
BPAQ-Anger
-0.61 (0.71)
-0.40 (0.71)
-2.37
.02
-0.30
-0.87 (0.56)
-1.09 (0.58)
2.90
.004
0.39
BPAQ-Hostility
-0.58 (0.70)
-0.49 (0.68)
-1.02
.31
-0.13
-1.13 (0.51)
-1.29 (0.52)
2.41
.02
0.31
BPAQ-Total
-0.59 (0.50)
-0.62 (0.45)
0.46
.65
0.06
-0.69 (0.43)
-0.81 (0.42)
2.13
.03
0.28
Feeling thermometer (degrees)
25.54 (20.71)
26.33 (18.99)
-0.31
.76
-0.04
24.98 (21.48)
23.35 (18.50)
0.62
.54
0.08
First-line entries= Sample I (114 men, 130 women; for IAT: 95 and 107), second-line entries= Sample II (101 men, 132 women; for IAT: 79 and 120).
d
= Cohen’s
d
(average group difference
[male mean minus female mean] divided through square root of weighted mean of group variances). BPAQ= Buss-Perry Aggression Questionnaire (mean item responses).
P
values are two-
tailed
Pág. 382-
29/6/09
19:40
Página 385
subscales and between these and the feeling thermometer were
consistently positive (lower-right corner entries). Implicitly
measured aggression was largely dissociated from the explicit
aggression measures, one noteworthy exception being a highly
significant correlation between IAT effect and Anger scores
emerging only for men in Sample II.
As for sex-specific associations of finger-length measures with
implicit or explicit aggression measures (upper-right and lower-left
corner entries), crucially, out of 8 (finger-length measures)
×
7
(aggression measures)
×
2 (sexes)
×
2 (samples)= 224 computed
correlations only 4 were significant. Not a single of these
associations replicated across samples or would have survived
multiple-tests correction, and 3 of them were directionally opposite
to theory, with the exception of a negative correlation between
R2D:4D and Physical aggression for men in Sample II (
r
= -.23,
p
<.05). Overall, the proportion of significant correlations (4 / 224=
1.8%) was even below the expected error rate of statistical tests
(5%, if
p
<.05), when the null hypothesis (no effect) actually is true.
Rerunning the correlational analysis using Spearman’s rank-order
correlation coefficient also failed to yield any discernible pattern of
consistent, replicable, or theory compliant associations of digit
ratios with explicit or implicit aggression measures (details
omitted), thus suggesting that the nil findings of the main analysis
were robust and not influenced by outliers.
Discussion
This study was only the second one to examine associations of
2D:4D with a candidate trait measured both explicitly and
implicitly. Samples I and II combined were almost six times larger
than in the predecessor study (Schmukle et al., 2007). Based on a
subsample of 39 men, the predecessor reported a significant
correlation of
r
= .33 between R2D:4D and an implicit (IAT-based)
measure of gender self-concept (i.e., masculinity-femininity).
Assuming that this correlation magnitudinally was typical (i.e.,
that associations between 2D:4D and implicitly measured target
MARTIN VORACEK AND STEFAN STIEGER
386
Table 3
Sex-specific associations among variables under study
R2D:4D
L2D:4D
D
R-L
R2D
R4D
L2D
L4D
Average
IAT
Physical
Verbal
Anger
Hostility
Total
Feeling
R2D:4D
.92***
.81***
.31**
.20*
-.39***
.15
-.32**
-.11
.04
-.04
.01
.10
.06
.04
.02
.83***
.50***
.58***
.37***
-.18‡
.23*
.02
.12
.03
-.23*
-.01
-.05
.01
-.10
-.06
L2D:4D
.66***
.89***
-.30**
.20*
-.29**
.26**
-.31**
-.05
.14
.02
-.07
.09
.14
.08
-.03
.49***
.86***
-.42***
.20*
-.07
.22*
-.21*
.04
.09
-.11
.05
-.15
-.04
-.08
.05
D
R-L
.50***
-.33***
.75***
.01
-.17‡
-.18‡
-.01
-.10
-.17‡
-.11
.14
.02
-.14
-.06
.09
.45***
-.57***
.71***
.20*
-.12
.03
.21*
.09
-.07
-.13
-.06
.08
.05
-.02
-.11
R2D
.13
.13
.02
.99***
.82***
.86***
.73***
.92***
.03
-.08
.09
-.01
-.01
-.02
-.03
.17*
-.01
.17*
.98***
.85***
.80***
.72***
.91***
.01
-.19‡
-.08
-.12
-.02
-.13
-.05
R4D
-.36***
-.20*
-.23**
.88***
.99***
.72***
.88***
.93***
.01
-.05
.08
-.06
-.03
-.02
-.03
-.27**
-.22*
-.02
.90***
.98***
.72***
.75***
.89***
-.01
-.08
-.07
-.10
-.03
-.09
-.02
L2D
-.01
-.16‡
-.18*
.95***
.89***
.99***
.84***
.92***
.08
-.06
.01
.03
.11
.03
-.01
.03
.06
-.03
.74***
.71***
.99***
.91***
.94***
-.05
-.11
-.03
-.10
-.03
-.09
-.02
L4D
-.29**
-.29**
-.04
.86***
.95***
.90***
.99***
.94***
-.01
-.07
.05
-.01
.03
-.01
.02
-.20*
.41***
.23**
.70***
.75***
.89***
.99***
.93***
-.09
-.06
-.05
-.04
-.01
-.05
-.04
Average
-.15‡
-.06
-.11
.95***
.97***
.97***
.97***
.99***
.03
-.07
.06
-.01
.03
-.01
-.01
-.08
.17*
.10
.91***
.92***
.91***
.91***
.99***
-.05
-.12
-.06
-.10
-.02
-.10
-.04
IAT
-.20*
-.10
-.13
-.08
.02
-.01
.04
-.01
.95
.03
-.10
-.05
.01
-.03
-.01
-.15
-.04
-.10
-.09
-.02
-.03
-.01
-.04
.96
.04
-.06
-.43***
-.06
-.16
-.07
Physical
.05
.09
-.05
-.05
-.06
-.05
-.09
-.06
.05
.77
.20*
.46***
.27**
.74***
.48***
.07
-.02
.08
-.06
-.09
.02
.03
-.03
.12
.62
.43***
.50***
.42***
.78***
.29**
Verbal
-.10
.04
-.17‡
-.05
.01
.01
-.01
-.01
-.07
.28**
.50
.34***
.20*
.51***
.40***
.01
-.01
.01
-.03
-.04
.01
.01
-.02
-.06
.40***
.53
.44***
.43***
.72***
.28**
Anger
-.02
.09
-.13
-.08
-.07
-.05
-.09
-.07
.08
.42***
.41***
.72
.46***
.79***
.48***
-.06
-.11
.06
-.17*
-.14
-.12
-.05
-.13
.16
.57***
.24**
.58
.56***
.80***
.27**
Hostility
.08
.08
-.01
-.07
-.10
-.07
-.10
-.09
.09
.30**
.17‡
.40***
.72
.73***
.30**
.08
.06
.02
-.09
-.13
-.08
-.10
-.11
.16‡
.48***
.28**
.50***
.63
.79***
.31**
Total
.02
.11
-.10
-.09
-.09
-.06
-.11
-.09
.07
.72***
.56***
.78***
.74***
.82
.58***
.04
-.02
.06
-.12
-.13
-.06
-.05
-.10
.14‡
.84***
.60***
.77***
.78***
.82
.38***
Feeling
.06
.19*
-.14
.04
.01
.06
-.02
.02
.08
.54***
.22*
.32***
.19*
.45***
.13
.03
.10
-.05
-.11
-.05
-.06
-.07
-.06
.43***
.10
.40***
.37***
.44***
First-line entries= Sample I, second-line entries= Sample II. Diagonal entries=
ICC
s (finger-length measurements) or Cronbach’s
α
(aggression measures). Off-diagonal entries= Pearson
correlations, above-diagonal entries= men, below-diagonal entries= women. ‡
p
<.10, *
p
<.05, **
p
<.01, ***
p
<.001 (two-tailed)
Pág. 382-
29/6/09
19:40
Página 386
REPLICATED NIL ASSOCIATIONS OF DIGIT RATIO (2D:4D) AND ABSOLUTE FINGER LENGTHS WITH IMPLICIT AND EXPLICIT MEASURES OF AGGRESSION
387
traits typically would yield
r
figures of about .30), the a priori
estimate of statistical power (Elashoff, 2000) to detect this
theoretical
r
value of .30 would have merely been 45% with two-
sided testing and a sample size of 39. This suggests that the finding
of Schmukle et al. (2007) may well be a chance finding. However,
given sample sizes of 79, 95, 107, or 120 (as were available for the
tests for within-sex associations of digit ratios with the IAT in
Samples I and II of the present study; see Table 1), the a priori
power estimates to detect an expected
r
value of .30 with two-
sided testing amount to 76%, 84%, 88%, or 91%, respectively.
These values of statistical power are sufficiently high and thus
suggest that the present two studies were not statistically
underpowered and that their nil findings are unlikely to be due to
statistical type II errors.
The current findings are clear-cut. Overall, there was no
indication of theory compliant, reliable, or replicable within-sex
associations of putative markers for prenatal (2D:4D, D
R-L
) or
pubertal-adolescent (absolute finger length) androgen action with
either implicit or explicit measures of aggression. The lack of
reliable 2D:4D associations with BPAQ scores is in line with nil
findings of 2D:4D studies that also used the BPAQ (Austin et al.,
2002; Coyne et al., 2007). Some recent studies (Millet & Dewitte,
2007, 2008, 2009; van den Bergh & Dewitte, 2006) point to the
possibility that contextual cues may moderate 2D:4D effects,
particularly cues eliciting or priming aggressive versus non-
aggressive (altruistic and prosocial) behavior. By its design, this
eventuality could not be addressed in the present research. However,
it is worth mentioning that (similar to other strands of evidence from
2D:4D research) at present such context-dependent interaction
effects for associations of 2D:4D with target traits have only been
reported by one research group (and thus await independent
replication) and furthermore have only been tested within a limited
range of behavioral responses, namely largely for economic
decision behavior (which calls for the need to demonstrate
generalizability of such assumed interaction effects to a wider range
of behavior). In our study, 2D:4D was also not reliably associated
with aggression measured implicitly with a personalized IAT
variant. In this regard, our systematic replication (or generalizability
test) of the findings of Schmukle et al. (2007) failed: it is therefore
uncertain (and remains to be tested anew) whether 2D:4D really is
stronger or more reliably or anyway associated with sexually
differentiated (and thus conceivably hormonally influenced)
candidate individual difference variables, when these are assessed
implicitly rather than explicitly. Further, explicit and implicit
aggression measures themselves were largely dissociated, a finding
not uncommon in IAT research (Lane et al., 2007).
Diverse knowns of previous research replicated in both
samples, thus attesting to effect robustness, procedural validity,
and data typicality. For instance, measurements of finger length
and digit ratios were highly repeatable, internal consistency
figures of the IAT were high (see Mierke & Klauer, 2003) and
those of BPAQ subscales dovetailed to published normative data
for its German form. Sex differences in digit ratios were
significant, small-to-medium sized, somewhat more pronounced
for R2D:4D than for L2D:4D, noticeably smaller and
nonsignificant (but directionally theory compliant) for D
R-L
. Those
in absolute finger length were large, but invariably larger for 4D
than for 2D, affirming the position that it is men’s relatively longer
4D that accounts for the sex effect seen in 2D:4D. BPAQ subscales
were throughout positively interrelated and likewise throughout
positively correlated with a simple, single-item explicit aggression
measure of high face validity (feeling thermometer). The novel
personalized IAT variant for assessing aggression showed
potential to yield sex effects in the predicted direction, although
magnitudinally these were small.
In contrast, the failure of sex effects in the BPAQ to replicate
reliably and directionally consistent across samples suggests
psychometric shortcomings of this instrument’s internal structure
(see Herzberg, 2003), which also have become apparent in several
adaptations of the BPAQ in other languages (Dutch version:
Meesters, Muris, Bosma, Schouten, & Beuving, 1996; Japanese
version: Nakano, 2001) as well as in an alternative German
version of the BPAQ (von Collani & Werner, 2005). For instance,
in one of their samples, von Collani and Werner (2005) found
significant, but small, sex differences in Physical aggression
scores (men higher:
d
= 0.31) and Anger scores (women higher:
d
=
-0.25), whereas no significant sex differences in Verbal aggression
or Hostility scores. The sex difference in Physical aggression
scores replicated in another sample and was medium-sized (men
higher:
d
= 0.64), but the sex difference in Anger scores did not
replicate, and, on the other hand, there was a significant sex
difference in Verbal aggression scores (contrary to expectation,
men higher:
d
= 0.22) in this second sample, which had been absent
in the first sample. Also, the sex difference in total BPAQ scores
in the second sample was, although statistically significant,
magnitudinally modest (men higher:
d
= 0.21).
More generally, the current nil findings add to a growing
number of failures in 2D:4D research to replicate initially reported
findings from small-
N
, single-sample designs in subsequent
investigations providing large-scale or multi-sample evidence or
both (e.g., Medland, Loehlin, & Martin, 2008). Presently, the
majority of 2D:4D research findings is unreplicated. Time will
show which of these will turn out unreplicable. In this regard, the
combined approach of systematic replication attempts of prior
evidence, along with exact replication attempts of one’s own
research, as pursued here, appears beneficial and of general
interest for the 2D:4D research program. In tandem with large-
scale, multi-sample studies and meta-analyses, this should help
sorting out veridical findings from those due to chance, selective
publishing, or related phenomena. To achieve this goal, making
original nil findings and replication failures visible will be of
utmost importance. By the same token, one conceivable
implication of the current design is to adjust publication
acceptance criteria in this field, such that a successful replication
of 2D:4D associations with target traits must be presented.
A final proposal concerns the reporting of replication statistics.
This should assist in gauging the veridicality of findings. For
example, calculated
p
rep
values for the sex effect in digit ratios
observed in our samples (Table 2) ranged from 0.94 to 0.98,
indicating a high degree of trustworthiness. Indeed, the sex effect
in 2D:4D undoubtedly constitutes the most robust and replicable
2D:4D research finding so far. However,
p
rep
figures for observed
correlations between finger-length and aggression measures were
much lower (many of them < .50), representing a high degree of
uncertainty to see the same effect replicated in another research
report. Or in one’s own next sample, as we did.
Pág. 382-
29/6/09
19:40
Página 387
MARTIN VORACEK AND STEFAN STIEGER
388
tive ability and digit ratio.
Personality and Individual Differences, 33,
1115-1124.
Bailey, A.A., & Hurd, P.L. (2005). Finger length ratio (2D:4D) correlates
with physical aggression in men but not in women.
Biological Psy-
chology, 68,
215-222.
Beech, J.R., & Mackintosh, I.C. (2005). Do differences in sex hormones
affect handwriting style? Evidence from digit ratio and sex role identity
as determinants of the sex of handwriting.
Personality and Individual
Differences, 39,
459-468.
Benderlioglu, Z., & Nelson, R.J. (2004). Digit length ratios predict reac-
tive aggression in women, but not in men.
Hormones and Behavior, 46,
558-564.
Buss, A.H., & Perry, M. (1992). The Aggression Questionnaire.
Journal of
Personality and Social Psychology, 63,
452-459.
Coyne, S.M., Manning, J.T., Ringer, L., & Bailey, L. (2007). Directional
asymmetry (right-left differences) in digit ratio (2D:4D) predict indi-
rect aggression in women.
Personality and Individual Differences, 43,
865-872.
Csathó, A., Osváth, A., Bicsák, E., Karádi, K., Manning, J.T., & Kállai, J.
(2003). Sex role identity related to the ratio of second to fourth digit
length in women.
Biological Psychology, 62,
147-156.
Egloff, B., & Schmukle, S.C. (2002). Predictive validity of an implicit as-
sociation test for assessing anxiety.
Journal of Personality and Social
Psychology, 83,
1441-1455.
Elashoff, J.D. (2000).
nQuery Advisor
®
Version 4.0
[computer software].
Cork, Ireland: Statistical Solutions Ltd.
Evardone, M. (2006).
Sex differences in anxiety: Testing a prenatal androgen
hypothesis using behavioral and physiological markers.
Unpublished
thesis, Texas A&M University.
Fawcett, C.C. (2003).
Individual difference factors and men’s attraction to
female body features.
Unpublished thesis, Brock University, Canada.
Fazio, R.H., & Olson, M.A. (2003). Implicit measures in social cognition
research: Their meaning and uses.
Annual Review of Psychology, 54
,
297-327.
Gallup, A.C., White, D.D., & Gallup, G.G., Jr. (2007). Handgrip strength
predicts sexual behavior, body morphology, and aggression in male
college students.
Evolution and Human Behavior, 28,
423-429.
Greenwald, A.G., McGhee, D.E., & Schwartz, J.L.K. (1998). Measuring
individual differences in implicit cognition: The Implicit Association
Test.
Journal of Personality and Social Psychology, 74
, 1464-1480.
Greenwald, A.G., & Nosek, B.A. (2001). Health of the Implicit Associati-
on Test at age 3.
Zeitschrift für Experimentelle Psychologie, 48
, 85-93.
Hampson, E., Ellis, C.L., & Tenk, C.M. (2008). On the relation between
2D:4D and sex-dimorphic personality traits.
Archives of Sexual Behavior,
37,
133-144.
Hendrick, C. (1990). Replications, strict replications and conceptual repli-
cations: Are they important? In J.W. Neuliep (Ed.):
Handbook of repli-
cation research in the behavioral and social sciences
(pp. 41-50). Corte
Madera, CA: Select Press.
Herzberg, P.Y. (2003). [Psychometric evaluation and validity of the Ger-
man translation of the Aggression Questionnaire by Buss and Perry].
Zeitschrift für Differentielle und Diagnostische Psychologie, 24
, 311-
323.
Hines, M. (2008). Early androgen influences on human neural and behavioural
development.
Early Human Development, 84,
805-807.
Jackson, C. (2008). Prediction of hemispheric asymmetry as measured by
handedness from digit length and 2D:4D digit ratio.
Laterality, 13,
34-
50.
Killeen, P.R. (2005). An alternative to null-hypothesis significance tests.
Psychological Science, 16,
345-353.
Kuepper, Y., & Hennig, J. (2007). Behavioral aggression is associated with
the 2D:4D ratio in men but not in women.
Journal of Individual Dif-
ferences, 28,
64-72.
Lane, K.A., Banaji, M.R., Nosek, B.A., & Greenwald, A.G. (2007). Un-
derstanding and using the Implicit Association Test: IV. What we know
(so far) about the method. In B. Wittenbrink & N. Schwarz (Eds.):
Im-
plicit measures of attitudes
(pp. 59-102). New York: Guilford Press.
Lippa, R.A. (2003). Are 2D:4D finger-length ratios related to sexual
Journal of Personality and
Social Psychology, 85,
179-188.
Lippa, R.A. (2006). Finger lengths, 2D:4D ratios, and their relation to gen-
der-related personality traits and the Big Five.
Biological Psychology,
71,
116-121.
Manning, J.T. (2002).
Digit ratio: A pointer to fertility, behavior, and
health.
New Brunswick, NJ: Rutgers University Press.
Manning, J.T., Churchill, A.J.G., & Peters, M. (2007). The effects of sex,
ethnicity, and sexual orientation on self-measured digit ratio (2D:4D).
Archives of Sexual Behavior, 36,
223-233.
Manning, J.T., Scutt, D., Wilson, J., & Lewis-Jones, D.I. (1998). The ratio
of 2nd to 4th digit length: A predictor of sperm numbers and concen-
trations of testosterone, luteinizing hormone and oestrogen.
Human Re-
production, 13,
3000-3004.
McFadden, D., Loehlin, J.C., Breedlove, S.M., Lippa, R.A., Manning, J.T.,
& Rahman, Q. (2005). A reanalysis of five studies on sexual orienta-
tion and the relative length of the 2nd and 4th fingers (the 2D:4D ratio).
Archives of Sexual Behavior, 34,
341-356.
McIntyre, M.H., Barrett, E.S., McDermott, R., Johnson, D.D.P., Cowden,
J., & Rosen, S.P. (2007). Finger length ratio (2D:4D) and sex differences
in aggression during a simulated war game.
Personality and Individual
Differences, 42,
755-764.
Medland, S.M., Loehlin, J.C., & Martin, N.G. (2008). No effects of prena-
tal hormone transfer on digit ratio in a large sample of same- and op-
posite-sex dizygotic twins.
Personality and Individual Differences, 44,
1225-1234.
Meesters, C., Muris, P., Bosma, H., Schouten, E., & Beuving, S. (1996).
Psychometric evaluation of the Dutch version of the Aggression Ques-
tionnaire.
Behaviour Research and Therapy, 34,
839-843.
Meingaßner, E. (2003). [
Somatic markers for in utero androgen levels and
spatial abilities
]
.
Unpublished thesis, University of Vienna.
Mierke, J., & Klauer, K.C. (2003). Method-specific variance in the Implicit
Association Test.
Journal of Personality and Social Psychology, 85
,
1180-1192.
Millet, K., & Dewitte, S. (2007). Digit ratio (2D:4D) moderates the impact
of an aggressive music video on aggression.
Personality and Individual
Differences, 43,
289-294.
Millet, K., & Dewitte, S. (2008). A subordinate status position increases
the present value of financial resources for low 2D:4D men.
American
Journal of Human Biology, 20,
110-115.
Millet, K., & Dewitte, S. (2009). The presence of aggression cues inverts
the relation between digit ratio (2D:4D) and pro-social behaviour in a
dictator game.
British Journal of Psychology, 100,
151-162.
Nakano, K. (2001). Psychometric evaluation on the Japanese adaptation of
the Aggression Questionnaire.
Behaviour Research and Therapy, 39,
853-858.
Nieschlag, E., Behre, H.M., & Nieschlag, S. (Eds.). (2004).
Testosterone:
Action, deficiency, substitution.
Cambridge, UK: Cambridge Universi-
ty Press.
Olson, M.A., & Fazio, R.H. (2004). Reducing the influence of extraper-
sonal associations on the Implicit Association Test: Personalizing the
IAT.
Journal of Personality and Social Psychology, 86
, 653-667.
Palmer, A.R. (2000). Quasi-replication and the contract of error: Lessons
from sex ratios, heritabilities and fluctuating asymmetry.
Annual Re-
view of Ecology and Systematics, 31,
441-480.
Puts, D.A., McDaniel, M.A., Jordan, C.L., & Breedlove, S.M. (2008). Spatial
ability and prenatal androgens: Meta-analyses of congenital adrenal
hyperplasia and digit ratio (2D:4D) studies.
Archives of Sexual Behavior,
37,
100-111.
Putz, D.A., Gaulin, S.J.C., Sporter, R.J., & McBurney, D.H. (2004). Sex
hormones and finger length: What does 2D:4D indicate?
Evolution and
Human Behavior, 25,
182-199.
Rammsayer, T.H., & Troche, S.J. (2007). Sexual dimorphism in second-to-
fourth digit ratio and its relation to gender-role orientation in males and
females.
Personality and Individual Differences, 42,
911-920.
Robinson, A.L. (2005).
Second to fourth digit ratio does not correlate with
dominance or masculinity in either males or females.
Unpublished the-
sis, Appalachian State University.
References
Pág. 382-
29/6/09
19:40
Página 388
REPLICATED NIL ASSOCIATIONS OF DIGIT RATIO (2D:4D) AND ABSOLUTE FINGER LENGTHS WITH IMPLICIT AND EXPLICIT MEASURES OF AGGRESSION
389
Scarbrough, P.S., & Johnston, V.S. (2005). Individual differences in
women’s facial preferences as a function of digit ratio and mental rota-
tion ability.
Evolution and Human Behavior, 26,
509-526.
Schicker, K. (2005). [
Digit ratio (2D:4D), sex-role typicality, normative
sex-role orientation, and sex-specific conflict behavior
]
.
Unpublished
thesis, University of Vienna.
Schmukle, S.C., Liesenfeld, S., Back, M.D., & Egloff, B. (2007). Second
to fourth digit ratios and the implicit gender self-concept.
Personality
and Individual Differences, 43,
1267-1277.
Troche, S.J., Weber, N., Hennigs, K., Andresen, C.-R., & Rammsayer, T.H.
(2007). The relationship of digit ratio (2D:4D) and gender-role orien-
tation in four national samples.
Journal of Individual Differences, 28,
78-87.
van den Bergh, B., & Dewitte, S. (2006). Digit ratio (2D:4D) moderates the
impact of sexual cues on men’s decisions in ultimatum games.
Pro-
ceedings of the Royal Society of London Series B, Biological Sciences,
273,
2091-2095.
von Collani, G., & Werner, R. (2005). Self-related and motivational
constructs as determinants of aggression: An analysis and validation of a
German version of the Buss-Perry Aggression Questionnaire.
Personality
and Individual Differences, 38,
1631-1643.
Vonnahme, P. (2005).
Facial preferences of Caucasian males as a function
of 2D:4D digit ratio and mental rotation ability.
Unpublished disserta-
tion, New Mexico State University.
Voracek, M., Dressler, S.G., & Manning, J.T. (2007). Evidence for assor-
tative mating on digit ratio (2D:4D), a biomarker for prenatal androgen
exposure.
Journal of Biosocial Science, 39,
599-612.
Voracek, M., Manning, J.T., & Dressler, S.G. (2007). Repeatability and in-
terobserver error of digit ratio (2D:4D) measurements made by experts.
American Journal of Human Biology, 19,
142-146.
Pág. 382-
29/6/09
19:40
Página 389
logo_pie_uaemex.mx